llvmpipe disables denorms on purpose (on x86/sse only), because denorms are
generally neither required nor desired for graphic apis (and in case of d3d10,
they are forbidden).
However, this caused some arithmetic tests using denorms to fail on some
systems, because the reference did not generate the same results anymore.
(It did not fail on all systems - behavior of these math functions is sort
of undefined when called with non-standard floating point mode, hence the
result differing depending on implementation and in particular the sse
capabilities.)
So, for the reference, simply flush all (input/output) denorms manually
to zero in this case.
This fixes https://bugs.freedesktop.org/show_bug.cgi?id=67672.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
(cherry picked from commit 8148a06b8f)
Nominated-by: Matt Turner <mattst88@gmail.com>
Back to the original commit (8313f444) adding the workaround, we were
enabling it on gens <= 7, even though gens <= 5 can't do multisampling.
I cannot find documentation that says that Sandybridge needs this
workaround but in practice disabling it causes these piglit tests to
fail:
EXT_framebuffer_multisample/interpolation {2,4} centroid-deriv{,-disabled}
On Ironlake:
total instructions in shared programs: 4358478 -> 4349671 (-0.20%)
instructions in affected programs: 117680 -> 108873 (-7.48%)
A bunch of shaders in TF2, Portal 2, and L4D2 are cut by 25~30%.
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
(cherry picked from commit 1a2de7dce8)
On evergreen there are 4 regs, on r600/700 there is only one.
Don't initialise regs and trash someone elses state.
Not sure this fixes anything, but hey one less stupid.
Reviewed-By: Glenn Kennard <glenn.kennard@gmail.com>
Cc: "10.3 10.4" mesa-stable@lists.freedesktop.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
(cherry picked from commit 7f21cf7198)
Nowadays GCC assumes stack pointer is 16-byte aligned even on 32-bits, but that is an assumption OpenGL drivers (or any dynamic library for that matter) can't afford to make as there are many closed- and open- source application binaries out there that only assume 4-byte stack alignment.
V4: fix comment and indentation
V3: move all sse4.1 build flag config to the same location
and add comment as to why we need to do the realign
V2: use $target_cpu rather than $host_cpu
and setup build flags in config rather than makefile
https://bugs.freedesktop.org/show_bug.cgi?id=86788
Signed-off-by: Timothy Arceri <t_arceri@yahoo.com.au>
Reviewed-by: Matt Turner <mattst88@gmail.com>
CC: "10.4" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit f1b5f2b157)
Harvested GPUs have some of their render backends disabled, so
in order to prevent the hardware from trying to render things
with these disabled backends we need to correctly program
the PA_SC_RASTER_CONFIG register.
v2:
- Write RASTER_CONFIG for all SEs.
v3:
- Set GRBM_GFX_INDEX.INSTANCE_BROADCAST_WRITES bit.
- Set GRBM_GFX_INFEX.SH_BROADCAST_WRITES bit when done setting
PA_SC_RASTER_CONFIG.
- Get num_se and num_sh_per_se from kernel.
v4:
- Get correct value for num_se
- Remove loop for setting PA_SC_RASTER_CONFIG
- Only compute raster config when a backend has been disabled.
v5: Michel Dänzer
- Fix computation for chips with multiple SEs
https://bugs.freedesktop.org/show_bug.cgi?id=60879
CC: "10.4 10.3" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 67dcbcd92c)
Fixes an infinite loop in swrast where the lowering pass unpacks saturate into
clamp but the opt_algebraic pass tries to do the opposite.
v3 (Ian):
This is a revert of commit cfa8c1cb "ir_to_mesa: lower ir_unop_saturate" on
the ir_to_mesa.cpp portion. prog_execute.c can handle saturates in vertex
shaders, so classic swrast shouldn't need this lowering pass.
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83463
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
(cherry picked from commit 39f7b72428)
This was an oversight in the original patch. When PolygonMode is
used, then front faces, back faces, or both may be rendered as
points and are affected by point sprite state.
Note that SNB/IVB can't actually be fully conformant here, for
a legacy context -- we don't have separate sets of pointsprite
enables for front and back faces. Haswell ignores pointsprite
state correctly in hardware for non-point rasterization, so can
do this correctly, but it doesn't seem worth it.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=86764
Reviewed-by: Matt Turner <mattst88@gmail.com>
(cherry picked from commit ed56c16820)
The GS has an interesting use for mul. Because the GS can emit multiple
vertices per input vertex, and it also has a unique count at the top of the URB
payload, the GS unit needs to be able to dynamically specify URB write offsets
(relative to the global offset). The documentation in the function has a very
good explanation from Paul on the mechanics.
This fixes around 2000 piglit tests on BSW.
v2:
Reworded commit message (Ben) no mention of CHV (Matt)
Change SHRT_MAX to USHRT_MAX (Ken, and Matt)
Update comment in code to reflect the use of UW (Ben)
Add Gen7+ assertion for the relevant GS code, since it won't work on Gen6- (Ken)
Drop the bogus hunk in emit_control_data_bits() (Ken)
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84777 (with many dupes)
Cc: "10.4 10.3 10.2" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
(cherry picked from commit f13870db09)
Nine code to match vertex declaration to vs inputs was limiting
the number of possible combinations.
Some sm3 games have issues with that, because arbitrary (usage/index)
can be used.
This patch does the following changes to fix the problem:
. Change the numbers given to (usage/index) combinations to uint16
. Do not put limits on the indices when it doesn't make sense
. change the conversion rule (usage/index) -> number to fit all combinations
. Instead of having a table usage_map mapping a (usage/index) number to
an input index, usage_map maps input indices to their (usage/index)
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Tested-by: Yaroslav Andrusyak <pontostroy@gmail.com>
Acked-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Axel Davy <axel.davy@ens.fr>
(cherry picked from commit 712a4c5438)
With sm3, you can declare an input/output with an usage and an usage index.
Nine code hardcodes the translation usage/index to a corresponding TGSI code.
The translation was limited to a few usage/index combinations that were corresponding
to most of the needs of games, but some games did not work.
This patch rewrites that Nine code to map all possible usage/index combination
to TGSI code. The index associated to TGSI_SEMANTIC_GENERIC doesn't need to be low
for good performance, as the old code was supposing, and is not particularly bounded
(it's UINT16). Given the index is BYTE, we can map all combinations.
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Tested-by: Yaroslav Andrusyak <pontostroy@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Axel Davy <axel.davy@ens.fr>
(cherry picked from commit 5d6d260833)
Issuing D3DISSUE_END should:
. reset previous queries if possible
. end the query
Previous behaviour wasn't calling end_query for
queries not needing D3DISSUE_BEGIN, nor resetting
previous queries.
This fixes several applications not launching properly.
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Tested-by: David Heidelberg <david@ixit.cz>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Axel Davy <axel.davy@ens.fr>
(cherry picked from commit eac0b9b68a)
Conflicts:
src/gallium/state_trackers/nine/query9.c
height=0 is legal for 1D array textures (as depth=0 is legal for
2D arrays). Fixes new piglit ext_texture_array-errors test.
Cc: "10.3 10.4" <mesa-stable@lists.freedesktop.org>
Reviewed-by: José Fonseca <jfonseca@vmware.com>
(cherry picked from commit 4e6244e80f)
Since 73dd50acf6
glsl: implement switch flow control using a loop
The SB backend was falling over in an assert or crashing.
Tracked this down to the loops having no repeats, but requiring
a working break, initial code just called the loop handler for
all non-if statements, but this caused a regression in
tests/shaders/dead-code-break-interaction.shader_test.
So I had to add further code to detect if all the departure
nodes are empty and avoid generating an empty loop for that case.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=86089
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Reviewed-By: Glenn Kennard <glenn.kennard@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
(cherry picked from commit 7b0067d23a)
PIPE_CAP_VIDEO_MEMORY returns the amount of video memory in megabytes,
so need to converted it to bytes.
Fixed Warframe memory detection.
v2: also prepare for cards with more than 4GB memory
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Tested-by: Yaroslav Andrusyak <pontostroy@gmail.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: David Heidelberg <david@ixit.cz>
(cherry picked from commit a99f31bced)
D3DPOOL_SCRATCH is disallowed according to spec.
D3DPOOL_SYSTEMMEM should be allowed but we don't handle it right for now.
v2: Fixes segfault in SetTexture when unsetting the texture
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Tested-by: David Heidelberg <david@ixit.cz>
Signed-off-by: Axel Davy <axel.davy@ens.fr>
(cherry picked from commit 4eea2496bc)
Fixes "Error : CONST[20]: Undeclared source register" when running
dx9_alpha_blending_material. Also artifacts on ilo.
v2: also remove unused MISC_CONST
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Tested-by: David Heidelberg <david@ixit.cz>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Axel Davy <axel.davy@ens.fr>
(cherry picked from commit 890f963d64)
This patch moves the data field from Resource9 to Surface9 and cleans
D3DPOOL_SYSTEMMEM handling in Texture9. This fixes HL2 lost coast.
It also removes in Texture9 some code written to support importing
and exporting non D3DPOOL_SYSTEMMEM shared buffers. This code hadn't
the design required to support the feature and wasn't used.
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Tested-by: David Heidelberg <david@ixit.cz>
Signed-off-by: Axel Davy <axel.davy@ens.fr>
(cherry picked from commit 6aeae7442d)
We are using 1 more buffer than we have, although in the future the
driver should just end up using one buffer in total probably, this
is a good first step, it merges the txq cube array and buffer info
constants on r600 and evergreen.
This should in theory fix geom shader tests on r600.
v1.1: fix comments from Glenn.
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
Cc: "10.4 10.3" <mesa-stable@lists.freedesktop.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
(cherry picked from commit 07ae69753c)
Squashed with commit
r600g: fix fallout from last patch
I accidentally rebased from the wrong machine and missed some
fixes that were on my r600 box.
doh.
this fixes a bunch of geom shader textureSize tests on rv635
from gpu reset to pass.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=86760
Reported-by: wolput@onsneteindhoven.nl
Cc: "10.4 10.3" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
(cherry picked from commit b10ddf962f)
Squashed with commit
r600g: make llvm code compile this time
Actually compiling the code helps make it compile.
Cc: "10.3 10.4" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
(cherry picked from commit 91a827624c)
It's not exported by the official opengl32.dll neither. Applications are
supposed to get it via wglGetProcAddress(), not GetProcAddress().
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
(cherry picked from commit cb009bdd44)
This fixes several MSVC warnings like:
warning C4273: 'glClearColorx' : inconsistent dll linkage
In fact, we should avoid using `declspec(dllexport)` altogether, and use
exclusively the .DEF instead, which gives more precise control of which
symbols must be exported, but all the public GL/GLES headers practically
force us to pick between `declspec(dllexport)` or
`declspec(dllimport)`.
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
(cherry picked from commit 5fdb6d6839)
Fixes broken rendering in Windows-based QtQuick2 apps run through Wine.
This library sets all texture units' GL_COORD_REPLACE, leaves point
sprite mode enabled, and then draws a triangle fan.
Will need a slightly different fix for Gen4-5, but I don't have my old
machines in a usable state currently.
V2: - Simplify patch -- the real changes are no longer duplicated across
the Gen6 and Gen7 atoms.
- Also don't clobber attr overrides -- which matters on Haswell too,
and fixes the other half of the problem
- Fix newly-introduced warnings
V3: - Use BRW_NEW_GEOMETRY_PROGRAM and brw->geometry_program rather than
core flag and state; keep the state flags in order.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84651
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
(cherry picked from commit 0008d0e59e)
Ilia noticed that my lowering pass was converting the constant array
used by textureGatherOffsets' offsets parameter to a uniform. This
broke textureGather for Nouveau, and is generally a horrible plan,
since it violates the GLSL constraint that offsets must be an
immediate constant.
When I wrote this pass, I neglected to consider whole array assignment.
I figured opt_array_splitting would handle constant indexing, so this
pass was really about fixing variable indexing.
textureGatherOffsets is an example of whole array access that we really
don't want to touch. Whole array copies don't appear to benefit from
this either - they're most likely initializers for temporary arrays
which are going to be mutated anyway. Since you're copying, you may
as well copy from immediates, not uniforms.
This patch makes the pass look for ir_dereference_arrays of
ir_constants, rather than looking for any ir_constant directly.
This way, it ignores whole array assignment.
No shader-db changes or Piglit regressions on Haswell. Some Piglit
tests generate different code (fixing textureGatherOffsets on Nouveau).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Tested-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 60f011af1a)
Uniform names (even for hidden uniforms) are required to be unique; some
parts of the compiler assume they can be looked up by name.
Fixes the piglit test: tests/spec/glsl-1.20/linker/array-initializers-1
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Cc: "10.4" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
(cherry picked from commit 129178893b)
When converting a uniform array reference to a pull constant load, the
`reladdr` expression itself may have its own `reladdr`, arbitrarily
deeply. This arises from expressions like:
a[b[x]] where a, b are uniform arrays (or lowered const arrays),
and x is not a constant.
Just iterate the lowering to pull constants until we stop seeing these
nested. For most shaders, there will be only one pass through this loop.
Fixes the piglit test:
tests/spec/glsl-1.20/linker/double-indirect-1.shader_test
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Cc: "10.3 10.4" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
(cherry picked from commit adefccd12a)
This moves all the CUBE section above the gradients section,
so that the gradient emission happens on one block which
is what sb/hardware expect.
v2: avoid changes to bytecode by using spare temps
v2.1: shame gcc, oh the shame. (uninit var warnings)
Cc: "10.4 10.3" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
(cherry picked from commit c88385603a)
The piglit tests were failing, and it appeared to be SB
optimising out things, but Glenn pointed out the gradients
are meant to be clause local, so we should emit the texture
instructions in the same clause. This moves things around
to always copy to a temp and then emit the texture clauses
for H/V.
v2: Glenn pointed out we could get another ALU fetch in
the wrong place, so load the src gpr earlier as well.
Fixes at least:
./bin/tex-miplevel-selection textureGrad 2D
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
Cc: "10.4 10.3" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
(cherry picked from commit 38ec184419)
res->bind is not an indicator of how the resource is currently bound.
buffers can be rebound across different binding points without changing
underlying storage.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.4 10.3" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit fecae4625c)
For 1D and 2D arrays we don't want the other coordinates being
offset and affecting where we sample. I wrote this patch 6 months
ago but lost it.
Fixes:
./bin/tex-miplevel-selection textureLodOffset 1DArray
./bin/tex-miplevel-selection textureLodOffset 2DArray
./bin/tex-miplevel-selection textureOffset 1DArray
./bin/tex-miplevel-selection textureOffset 1DArrayShadow
./bin/tex-miplevel-selection textureOffset 2DArray
./bin/tex-miplevel-selection textureOffset(bias) 1DArray
./bin/tex-miplevel-selection textureOffset(bias) 2DArray
v2: rewrite to handle more cases and be consistent with code
above.
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
Cc: "10.3 10.4" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
(cherry picked from commit 1830138cc0)
Not all drivers can set gl_Layer from VS. Add a fallback that passes the
instance id from VS to GS, and then uses the GS to set the layer.
Tested by adding
quad_buffers |= clear_buffers;
clear_buffers = 0;
to the st_Clear logic, and forcing set_vertex_shader_layered in all
cases. No piglit regressions (on piglits with 'clear' in the name).
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Cc: "10.4 10.3" <mesa-stable@lists.freedesktop.org>
(cherry picked from commit 68db29c434)
Some of the geom shader tests produce an empty vertex shader,
on cayman we'd crash in the finaliser because last_cf was NULL.
cayman doesn't need the NOP workaround, so if the code arrives
here with no last_cf, just emit an END.
fixes crashes in a bunch of piglit geom shader tests.
Cc: "10.3 10.4" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
(cherry picked from commit 4e520101e6)
The sampler_array_size field was added by "mesa/st: add support for
dynamic sampler offsets". But the field wasn't getting copied in
the get_pixel_transfer_visitor() or get_bitmap_visitor() functions.
The count_resources() function then didn't properly compute the
glsl_to_tgsi_visitor::samplers_used bitmask. Then, we didn't declare
all the sampler registers in st_translate_program(). Finally, we
asserted when we tried to emit a tgsi ureg src register with File =
TGSI_FILE_UNDEFINED.
Add the missing assignments and some new assertions to catch the
invalid register sooner.
Cc: "10.3, 10.4" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
(cherry picked from commit 11abd7b2bc)
Using the asynchronous DMA engine for multi-dimensional operations seems
to cause random GPU lockups for various people. While the root cause for
this might need to be fixed in the kernel, let's disable it for now.
Before re-enabling this, please make sure you can hit all newly enabled
paths in your testing, preferably with both piglit and real world apps,
and get in touch with people on the bug reports below for stability
testing.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=85647
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83500
Cc: "10.3 10.4" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Grigori Goronzy <greg@chown.ath.cx>
(cherry picked from commit ae4536b4f7)
So when checking/building sse code we have three possibilities:
1 Old compiler, throws an error when using -msse*
2 New compiler, user disables sse* (-mno-sse*)
3 New compiler, user doesn't disable sse
The original code, added code for #1 but not #2. Later on we patched
around the lack of handling #2 by wrapping the code in __SSE4_1__.
Yet it lead to a missing/undefined symbol in case of #1 or #2, which
might cause an issue for #2 when using the i965 driver.
A bit later we "fixed" the undefined symbol by using #1, rather than
updating it to handle #2. With this commit we set things straight :)
To top it all up, conventions state that in case of conflicting
(-enable-foo -disable-foo) options, the latter one takes precedence.
Thus we need to make sure to prepend -msse4.1 to CFLAGS in our test.
v2: Clean the #includes. Suggested by Ilia, Matt & Siavash.
Cc: "10.3 10.4" <mesa-stable@lists.freedesktop.org>
Tested-by: David Heidelberg <david@ixit.cz>
Tested-by: Siavash Eliasi <siavashserver@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
(cherry picked from commit 1a6ae84041)
DRI_PRIME setups have different issues due the lack of dma-buf fences
support in the drivers. For DRI3 DRI_PRIME, a race can appear, making
tearings visible, or worse showing older content than expected. Until
dma-buf fences are well supported (and by all drivers), an alternative
is to send the buffers to the server only when rendering has finished.
Since waiting the rendering has finished in the main thread has a
performance impact, this patch uses an additional thread to offload the
wait and the sending of the buffers to the server.
Acked-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: David Heidelberg <david@ixit.cz>
Signed-off-by: Axel Davy <axel.davy@ens.fr>
(cherry picked from commit 7f565845a1)
Implements vblank_mode and throttling, which allows us change default ratio
between framerate and input lag.
Acked-by: Jose Fonseca <jfonseca@vmware.com>
Signed-off-by: David Heidelberg <david@ixit.cz>
Signed-off-by: Axel Davy <axel.davy@ens.fr>
(cherry picked from commit 948e6c5228)
Work of Joakim Sindholt (zhasha) and Christoph Bumiller (chrisbmr).
DRI3 port done by Axel Davy (mannerov).
v2: - nine_debug.c: klass extended from 32 chars to 96 (for sure) by glennk
- Nine improvements by Axel Davy (which also fixed some wine tests)
- by Emil Velikov:
- convert to static/shared drivers
- Sort and cleanup the includes
- Use AM_CPPFLAGS for the defines
- Add the linker garbage collector
- Restrict the exported symbols (think llvm)
v3: - small nine fixes
- build system improvements by Emil Velikov
v4: [Emil Velikov]
- Do no link against libudev. No longer needed.
Acked-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Signed-off-by: David Heidelberg <david@ixit.cz>
(cherry picked from commit fdd96578ef)
[Emil Velikov: use correct ureg_property* functions]
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
v3: thanks to Brian, improved coding style, also glennk helped spot few
things (unsigned -> int, two constify)
v4: thanks Ilia improved function, dropped u_box_clip_3d
v5: incorporated rest of Gregor proposed changes,clean ups
v6: u_box_clip_2d simplify proposed by Ilia Mirkin
Acked-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: David Heidelberg <david@ixit.cz>
(cherry picked from commit 7d2573b537)
At this moment we use only zero or positive values.
v2: Implement it for also for Solaris, MSVC assembly
and enable for other combinations.
v3: Replace MSVC assembly by assert + warning during compilation
v4: remove inc and dec with return for MSVC assembly
Acked-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: David Heidelberg <david@ixit.cz>
(cherry picked from commit cb49132166)
Implement pipe_loader_sw_probe_wrapped which allows to use the wrapped
software renderer backend when using the pipe loader.
v2: - remove unneeded ifdef
- use GALLIUM_PIPE_LOADER_WINSYS_LIBS
- check for CALLOC_STRUCT
thanks to Emil Velikov
Acked-by: Jose Fonseca <jfonseca@vmware.com>
Signed-off-by: David Heidelberg <david@ixit.cz>
(cherry picked from commit e23d63cffd)
This will be reused for the scalar VS pass.
v2 (Ken): Rebase on master.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We'll reuse this toplevel optimization driver for the scalar VS.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
These last few operations all only apply when we've actually generated
code, optimized and allocated registers. The dummy and the repclear
shaders don't need the gen4 send workaround, and don't spill. This
means we can move these lines into the else-branch, which will make
the following refactoring easier.
v2 (Ken): Rebase on master, which removed the uncompressed stack.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We split out SIMD8 and SIMD16 generation into seperate calls to
new method generate_code(), which returns the start offset for the
generated code. A new get_assembly() method returns the generated code.
This avoids asserting MESA_SHADER_FRAGMENT and accessing wm_prog_data
in the generator.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Derived from st/glx's GLX_EXT_create_context_es/es2_profile implementation.
Tested with an OpenGL ES 2.0 ApiTrace.
Reviewed-by: Brian Paul <brianp@vmware.com>
The latest version of the specs explicitly allow it, and given that Mesa
universally supports KHR_debug we should definitely support it.
Totally untested. (Just happened to noticed this while implementing
GLX_EXT_create_context_es2_profile for st/xlib.)
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
17 insertions(+), 102 deletions(-). Works just as well.
v2: Make emit_math take const references (suggested by Matt),
drop redundant WRITEMASK_XYZW setting (Matt and Curro).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Every other unit in the geometry pipeline automatically enables
statistics gathering. This part of the pipe has been controlled by the
DEBUG_STATS variable, but this is asymmetric. This dates back to the
original implementation, and I am not sure if there is a reason for it.
I need access to these stats to implement ARB_pipeline_statistics_query.
Eric wrote it, and Ken touched it last. Do you have any opposition?
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=86145
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
According to gen2 BSpec the pipeline must be flushed at least up to the
windower before changing the scissor rect enable field. Emitting the
3DSTATE_SCISSOR_RECTANGLE_0 before 3DSTATE_SCISSOR_ENABLE is sufficient
to do that.
gen3 BSpec no longer has that piece of text, but let's make the same
change there too for symmetry. The spec does still say that the scissor
rectangle must be defined before enabling it, so the new order does seem
more in line with the spec.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Gen2 doesn't have fragment shaders so we shouldn't be calling
_mesa_meta_glsl_Clear() on gen2. Restore the appropriate
ARB_fragment_shader check to the clear path which was lost in:
commit 94f22fbe78
Author: Tapani Pälli <tapani.palli@intel.com>
Date: Wed Aug 8 20:46:45 2012 +0300
intel: use _mesa_meta_Clear with OpenGL ES 1.1 v2
v2: Fix spelling in commit message
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
TEXTURE_SET() is the only register macro that forgets to wrap the
argument evaluation in parens. Only simple integers are passed to this
macro so there's no bug but sitll it seems prudent to add the
parens.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Gen2 doesn't support depth/stencil textures, and since
commit c1d4d49993
Author: Ville Syrjälä <ville.syrjala@linux.intel.com>
Date: Thu Apr 24 14:11:43 2014 +0300
i915: Don't advertise Z formats in TextureFormatSupported on gen2
depth/stencil formats are no longer accepted as texture formats.
However we still want depth/stencil renderbuffers, so add explicit
format checks to intel_alloc_renderbuffer_storage() to allow such
things.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
gen2 doesn't supporte linear mip filter with anisotropic min/mag
filtering. The hardware would automagically downgrade the min/mag
filters to linear in such cases, which IMO looks worse than forcing
the mip filter to nearest.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
The spec says using DOT4 for alpha is undefined unless DOT4 is also used
for color. It seems to do the right thing anyway, but better safe than sorry.
Also override numAlphaArgs to 2 for DOT4 since that's what it wants.
This migth fix something in case the specified alpha mode has only one
argument. Also avoids emitting a needless 3DSTATE_MAP_BLEND_ARG if
the specified alpha mode has three arguments.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
All the shaders we've received so far had this be the case, but with
nir-to-tgsi that changed.
I might decide to make nir-to-tgsi keep the outputs in the same order, for
debugging sanity, but I'm not sure.
apitrace now supports it, and it makes it much easier to test
tracing/replaying on OpenGL ES contexts since
GLX_EXT_create_context_{es2,es}_profile are widely available.
Reviewed-by: Brian Paul <brianp@vmware.com>
I used these in the SEL peephole, but they require extra tracking and
fix ups. The SEL peephole can pretty easily find the blocks it needs
without these.
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Make the helpers fill out valid Gen7 3DSTATE_SF and 3STATE_SBE. This
prevents the helpers from having to do
dw[0] = GEN7_SBE_DW1_x; // setting DW1 value to dw[0]!?
and simplifies gen7_3DSTATE_{SF,SBE}().
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Render stream and render enable are independent from so enable. Having a
single return point makes it easier to see that.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Started to make pipe_stream_output_info mandatory, but ended up adding support
for stream id and making a workaround Gen7-specific.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
It replaces gen6_fill_3dstate_constant(). gen6_3DSTATE_CONSTANT_{VS,GS,PS}
are made wrappers of the new function.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Add gen6_so_3DSTATE_GS(), gen6_disable_3DSTATE_GS(), and
gen7_disable_3DSTATE_GS() to do SO on GEN6 or to disable GS.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
On x86-64 this saves 8 bytes of padding in the structure, and this
reduces the size of the structure to 32 bytes.
v2: Fix constructor so that GCC won't warn about the order of
initialization.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Due to the total number of bits used in the bitfield, this does not
increase the size of the structure.
It does, however, reduce the number of instructions required each time
one of these fields is accessed. To access ::matrix_columns with the
bitfield, three instructions were required:
movzbl 0x9(%rdx),%eax
shr %al
and $0x7,%eax
As a uint8_t, only one instruction is required.
movzbl 0xa(%rdx),%eax
These fields are accessed *a lot*.
Valgrind callgrind results for a trace of Tesseract:
_mesa_Uniform4fv _mesa_Uniform4f _mesa_Uniform1i
Before (64-bit): 48,103,497 16,556,096 676,447
After (64-bit): 45,722,616 15,737,964 670,607
_mesa_Uniform4fv _mesa_Uniform4f _mesa_Uniform1i
Before (32-bit): 61,472,611 21,051,222 821,361
After (32-bit): 57,987,421 19,872,226 811,609
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
There are no uniforms in OpenGL ES 1.x, so we can't even get to this
code in that API.
Also, reorder the checks. First check that transpose is true, then
check whether or not that is legal in the current API. transpose should
never be true in an ES2 context, so this gets one check (the more
expensive one) out of the main path.
Valgrind callgrind results for a trace of Tesseract:
_mesa_UniformMatrix4fv _mesa_UniformMatrix3fv
Before (64-bit): 96,119,025 24,240,510
After (64-bit): 90,726,569 22,926,662
_mesa_UniformMatrix4fv _mesa_UniformMatrix3fv
Before (32-bit): 132,434,452 29,051,808
After (32-bit): 126,658,112 27,989,316
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Before ARB_explicit_uniform_location, Mesa's location encoding allowed
locations for non-array types that had non-zero array indices.
Basically, part of the location was the uniform and part was the array
index. This meant that some checks had to occur for arrays and
non-arrays. This is no longer possible, we the checks can be split up.
Valgrind callgrind results for a trace of Tesseract:
_mesa_Uniform4fv _mesa_Uniform4f _mesa_Uniform1i
Before (64-bit): 50,499,557 17,487,316 686,227
After (64-bit): 50,023,791 17,274,432 684,293
_mesa_Uniform4fv _mesa_Uniform4f _mesa_Uniform1i
Before (32-bit): 62,968,039 21,732,380 828,147
After (32-bit): 62,373,967 21,490,756 826,223
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
I really wanted to remove 'shProg != NULL' as well, but that would have
required adding a dummy program as the default program. That seemed
like more churn than removing one test was worth.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Only one caller wanted to generate an error when location == -1, so move
the error generation to that caller. There will be more callers in the
future that do not want to generate errors.
Move the location == -1 check later in validate_uniform_parameters. As
currently implemented, glUniform1iv(-1, -1, data) would not generate an
error, but it should due to count being < 0.
The location that I have moved it to will make more sense with the next
commit.
Valgrind callgrind results for a trace of Tesseract:
_mesa_Uniform4fv _mesa_Uniform4f _mesa_Uniform1i
Before (64-bit): 51,241,217 17,740,162 689,181
After (64-bit): 50,499,557 17,487,316 686,227
_mesa_Uniform4fv _mesa_Uniform4f _mesa_Uniform1i
Before (32-bit): 63,940,605 21,987,918 831,065
After (32-bit): 62,968,039 21,732,380 828,147
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
The GL_ enums were previously used because glsl_types.h couldn't be used
in C code. That was fixed some time ago (and uniforms.c already
includes glsl_types.h), so this is no longer necessary.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Derive whether a RT supports blending, logicop, and the like when
set_framebuffer_state() is called. This enables us to simplify
gen6_BLEND_STATE().
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
According to the documentation, line widths higher than 40.0 may have
quality problems. That's already 20 times larger than we've been
exposing, so it seems totally sufficient.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
We've artificially been limiting this to 5 for no particular reason.
On Gen4-5, the limit is [0, 7.5] with a granularity of 0.5 (U3.1).
On Gen6+, the limit is [0, 7.9921875]. Since it's a U3.7, the
granularity should be 0.125 (1/8).
This patch conservatively advertises one granularity smaller than the
hardware's maximum value, just in case there's a problem using the
largest possible value. On Gen4-5, this is 7.5 - 0.5 = 7.0. On Gen6+,
this is 8.0 - 0.125 = 7.875.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Rather than hardcoding platform values in every code path, just use the
maximum value we set.
Currently, ctx->Const.LineWidth == 5, which is smaller than the hardware
limit. But applications shouldn't be using a value larger than we
support anyway.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Line Width moved to DW1 bits 29:12. It's actually now a U11.7.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Use clamping constants that guarantee no integer overflows.
As spotted by Chris Forbes.
This causes the code to change as:
- value |= (uint32_t)CLAMP(src[0], 0.0f, 4294967295.0f);
+ value |= (uint32_t)CLAMP(src[0], 0.0f, 4294967040.0f);
- value |= (uint32_t)((int32_t)CLAMP(src[0], -2147483648.0f, 2147483647.0f));
+ value |= (uint32_t)((int32_t)CLAMP(src[0], -2147483648.0f, 2147483520.0f));
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
On Windows, DllMain calls and thread creation/destruction are
serialized, so when llvmpipe is destroyed from DllMain waiting for the
rasterizer threads to finish will deadlock.
So, instead of waiting for rasterizer threads to have finished, simply wait for the
rasterizer threads to notify they are just about to finish.
Verified with this very simple program:
#include <windows.h>
int main() {
HMODULE hModule = LoadLibraryA("opengl32.dll");
FreeLibrary(hModule);
}
Fixes https://bugs.freedesktop.org/show_bug.cgi?id=76252
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Cc: 10.2 10.3 <mesa-stable@lists.freedesktop.org>
No longer needed as the file was removed with
commit 8c229d306b
Author: Kenneth Graunke <kenneth@whitecape.org>
Date: Mon Aug 11 10:07:07 2014 -0700
i965: Delete the Gen8 code generators.
We now use the brw_eu_emit.c code instead.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
During teardown we free the driver_configs list pointer, but we forget
to deallocate each config in that list.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-and-tested-by: Kenneth Graunke <kenneth@whitecape.org>
The function is not called by platform_drm. As such one needs to
pay special attention at teardown.
v2: Fix the comment block. Spotted by Ken.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-and-tested-by: Kenneth Graunke <kenneth@whitecape.org> (v1)
Earlier commit failed to attribure that for drm platforms one does not
call dri2_create_screen, thus it does not create the screen and
driver_configs but inherits them from the "display" - gbm.
As such wrap cleanup in Platform != _EGL_PLATFORM_DRM to prevent
the issue and still cleanup correctly for non-drm platforms.
v2:
- Drop the ifdef HAVE_DRM_PLATFORM, reindent the code and fix the
comment block. Suggested by Ken.
Reported-by: Kenneth Graunke <kenneth@whitecape.org>
Reported-by: Mark Janes <mark.a.janes@intel.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-and-tested-by: Kenneth Graunke <kenneth@whitecape.org> (v1)
Move opcode to string mappings to functions of their own. Have for consistent
outputs for similar opcodes.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
When the earlier block ended with control flow, we'd mistakenly remove
some of its links to its children. The same happened with the later
block.
Acked-by: Jason Ekstrand <jason.ekstrand@intel.com>
A pattern in certain shaders is:
uniform vec4 colors[NUM_LIGHTS];
for (int i = 0; i < NUM_LIGHTS; i++) {
...use colors[i]...
}
In this case, the application author expects the shader compiler to
unroll the loop. By doing so, it replaces variable indexing of the
array with constant indexing, which is more efficient.
This patch extends the heuristic to see if arrays accessed within the
loop are indexed by an induction variable, and if the array size exactly
matches the number of loop iterations. If so, the application author
probably intended us to unroll it. If not, we rely on the existing
loop-too-large heuristic.
Improves performance in a phong shading microbenchmark by 2.88x, and a
shadow mapping microbenchmark by 1.63x. Without variable indexing, we
can upload the small uniform arrays as push constants instead of pull
constants, avoiding shader memory access. Affects several games, but
doesn't appear to impact their performance.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Consider GLSL code such as:
const ivec2 offsets[] =
ivec2[](ivec2(-1, -1), ivec2(-1, 0), ivec2(-1, 1),
ivec2(0, -1), ivec2(0, 0), ivec2(0, 1),
ivec2(1, -1), ivec2(1, 0), ivec2(1, 1));
ivec2 offset = offsets[<non-constant expression>];
Both i965 and nv50 currently handle this very poorly. On i965, this
becomes a pile of MOVs to load the immediate constants into registers,
a pile of scratch writes to move the whole array to memory, and one
scratch read to actually access the value - effectively the same as if
it were a non-constant array.
We'd much rather upload large blocks of constant data as uniform data,
so drivers can simply upload the data via constbufs, and not have to
populate it via shader instructions.
This is currently non-optional because both i965 and nouveau benefit
from it, and according to Marek radeonsi would benefit today as well.
(According to Tom, radeonsi may want to handle this itself in the long
term, but we can always add a flag when it becomes useful.)
Improves performance in a terrain rendering microbenchmark by about 2x,
and cuts the number of instructions in about half. Helps a lot of
"Natural Selection 2" shaders, as well as one "HOARD" shader.
total instructions in shared programs: 5473459 -> 5471765 (-0.03%)
instructions in affected programs: 5880 -> 4186 (-28.81%)
v2: Use ir_var_hidden to avoid exposing the new uniform via the GL
uniform introspection API.
v3: Alphabetize Makefile.sources properly.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=77957
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
In the compiler, we'd like to generate implicit uniforms for internal
use. These should not be visible via the GL uniform introspection API.
To support that, we add a new ir_variable::how_declared value of
ir_var_hidden, and plumb that through to gl_uniform_storage.
v2 (idr): Fix some memory management issues in
move_hidden_uniforms_to_end. The comment block on the function has more
details.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Makes use of SSE 4.1 to speed up compute of min and max elements.
Callgrind cpu usage results from pts benchmarks:
Openarena 0.8.8: 3.67% -> 1.03%
UrbanTerror: 2.36% -> 0.81%
V5:
- actually make use of the optimisation in android (Emil Velikov)
- set a better array size limit for using SSE and added TODO
V4:
- fixed bugs with incrementing pointer and updating counters
V3:
- Removed sse_minmax.c from Makefile.sources
- handle the first few values without SSE until the pointer is aligned
and use _mm_load_si128 rather than _mm_loadu_si128
- guard the call to the SSE code better at build time
V2:
- removed GL* types
- use _mm_store_si128() rather than _mm_store_ps()
- add runtime check for SSE
- use aligned attribute for local mix/max
- bunch of tidyups
Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Timothy Arceri <t_arceri@yahoo.com.au>
Walk through the list and free each config, and finally free the list
itself. Freeing approx 20KiB of memory, according to valgrind.
Inspired by a similar patch by enpeng xu.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
driUnbindContext() checks for valid drawables before calling the driver
unbind function. In case of Surfaceless contexts, the drawables are always
Null and we end up not releasing the underlying DRI context. Moving the
call to the driver function before the drawable validity checks fixes things.
Steps to trigger this bug are following:
- create surfaceless context and make it current
- make some other context current
- {another thread} destroy surfaceless context
- make another context current
Signed-off-by: Alexandros Frantzis <Alexandros.Frantzis@canonical.com>
Signed-off-by: Kalyan Kondapally <kalyan.kondapally@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=74563
The dummy shader sends an EOT message to end itself. There are many more
works need to be done on the compiler side before we can advertise
PIPE_CAP_COMPUTE.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Use util_dynarray in ilo_set_global_binding() to allow for unlimited number of
global bindings. Add a comment for global bindings.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
We need to know the local/input/private sizes and others. This is not
complete. We need many others for CURBE setup.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
They will be used to report compute params or program compute states.
thread_count can also be used for 3DSTATE_VS.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
This started hitting an assertion recently. Only affects Haswell
(Ivybridge doesn't support this meddling with the sampler state pointer,
and ARB_gpu_shader5 is not enabled yet on Broadwell)
14 Piglits crash->pass.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Improves performance in GLBenchmark 2.7 TRex by 3.88889% +/- 0.336383%
(n=80) at 1280x720 on Broadwell GT3. Together with the previous patch,
it improves performance by 5.42738% +/- 0.541971% (n=10) at 1920x1080.
Note that without the PMA stall fix, this would instead decrease
performance by 22%.
v2: Update comment (noticed by Kristian Høgsberg).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Certain non-promoted depth cases typically incur stalls. In very
specific cases, we can enable a workaround which improves performance.
Improves performance in GLBenchmark 2.7 TRex by 1.17762% +/- 0.448765%
(n=75) at 1280x720 on Broadwell GT3.
Haswell has this feature as well, but we can't currently write registers
from userspace batches (and we'd incur additional software batch
scanning overhead as well), so we haven't enabled it. Broadwell allows
us to write CACHE_MODE_1. Backporters beware: the formula and flushing
incantation differs between Haswell and Broadwell.
v2: Move pma_stall_bits from brw->state to brw itself (requested by
Kristian Høgsberg).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Matt requested this in review feedback on the original patch, which I
completely missed when pushing this series. Kristian also made this
change, but I grabbed the wrong version of the patch.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Otherwise, calling glPopAttrib on drivers that don't support
ARB_clip_control gives you a GL error, which is surprising at best.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
We're not programming the clear values yet, so this won't work.
This patch should be (effectively) reverted eventually.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
On Skylake, the MOCS bits are an index into a table of 63 different,
configurable cache configurations. As for previous GENs, we only care about
WB and WT, which are available in the documented default set. Define
SKL_MOCS_WB and SKL_MOCS_WT to the indices for those configucations and use
those for the Skylake MOCS values.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Skylake has separate controls for enabling the Z Clip Test for the near
and far planes. For now, maintain the legacy behavior by setting both.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Skylake's 3DSTATE_DS packet has a few more fields; we don't support
domain shaders yet though.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
They are the same as for BDW, so just add a case for SKL to the init switch.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
On SKL, 3DSTATE_CONSTANT_* command is not committed until we give
the corresponding 3DSTATE_BINDING_TABLE_POINTERS_* command. If we
fail to do so, the constant buffers wont be read and push constants
will be wrong.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Otherwise they overlap and horrible things happen. All the new DWords
are for fast color clear values, which we don't do yet.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
We will need to allocate more DWords on Skylake.
v2: Don't mark brw_context parameter const. It's modified.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Skylake introduces a new base address for a feature we don't yet expose.
Setting these to 0 should be safe.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Skylake uploads the stencil reference values in DW3 of the
3DSTATE_WM_DEPTH_STENCIL packet, rather than in COLOR_CALC_STATE.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Skylake has some extra bits in PIPELINE_SELECT, none of which are
interesting for a 3D driver. In order to selectively change them, it
also introduces new "mask bits" in 15:8. We care about the "Pipeline
Selection" bits (1:0), so set the mask to 0x3.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
This commands has seen the addition of 2 dwords that allow to specify
which channels of which attributes need to be forwarded to the fragment
shader.
v2: Rebase forward a year (done by Ken).
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
The CSE pass now prints out why it thinks a value is not a candidate for
adding to the AE set.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
The optimization in commit d056863b covers these cases, which were the
first optimizations I added to the GLSL compiler.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Should trigger CL_INVALID_VALUE if device_list is NULL and num_devices
is greater than zero.
Introduced by e5468dfa52
Reported by: EdB
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Between release 3.2 and 3.3 LLVM stopped aligning properly when certain
conditions (no allocas, but large number of vectors causing spills to
the stack, and frame pointer omission enabled).
We were already disabling frame-pointer-omission on several build types,
but we now disable it on all build types.
It's not clear whether this affects 32-bits x86 processes only, or if it
can also affect 64-bits x86_64 processes when AVX registers are
available and used. So disable frame-pointer-omission on both
x86/x86_64 to be on the safe side.
See also:
- http://llvm.org/PR21435
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
AFAICT the number of threads is 80, not 70. I am not sure if Ken knows
something I do not.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Need to do a sqrt().
FWIW, the html that Sphinx 1.1.3 generates for the math expressions
looks completely broken.
Reviewed-by: José Fonseca <jfonseca@vmware.com>
Pass and return tgsi_token buffers instead of pipe_shader_state.
And update softpipe driver (the only user of this function).
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
Fixes polygon stipple if both DO_PSTIPPLE_IN_DRAW_MODULE and
DO_PSTIPPLE_IN_HELPER_MODULE are zero/off.
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
This was a regression introduced by
611d66fe45
Passing a binary program to clBuildProgram() is legal, but passing one
to clCompileProgram() is not.
v2:
- Code cleanups.
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
This adds a query which allows drivers to access the config
information of a specific function within the LLVM generated ELF
binary. This makes it possible for the driver to handle ELF
binaries with multiple kernels / global functions.
We are about to change mesa to spawn threads for deferred glCompileShader and
glLinkProgram, and we need to make sure those threads can send compiler
warnings/errors to the debug output safely.
Signed-off-by: Chia-I Wu <olv@lunarg.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
There may be two contexts compiling shaders at the same time, and we want the
anonymous struct id to be globally unique.
Signed-off-by: Chia-I Wu <olv@lunarg.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
_mesa_strtod and _mesa_strtof may be called from multiple threads. They need
to be thread-safe.
v2: platform checks are now done in configure.ac
Signed-off-by: Chia-I Wu <olv@lunarg.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
With the assumptions that xlocale.h implies newlocale and strtof_l. SCons is
updated to define HAVE_XLOCALE_H on linux and darwin.
Signed-off-by: Chia-I Wu <olv@lunarg.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Both core mesa and glsl have their own wrappers for strtof_l. Merge
and move them to util/. They are compiled with a C++ compiler so that
we can make them thread-safe in a following commit.
Signed-off-by: Chia-I Wu <olv@lunarg.com>
Reviewed-by: Kenneth Graunke <kenneth@whiteacpe.org>
This reverts commit cabc93c5ad.
Mark thinks the failures on the SNB GT2 in the lab are actually because
of faulty hardware, not instruction compaction. The GT1 didn't see any
problems after changes to the compaction code.
Helps a small number of vertex shaders in the games Dungeon Defenders
and Shank, as well as an internal benchmark.
instructions in affected programs: 2801 -> 2719 (-2.93%)
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
v2: Use the UST value provided in the PRESENT_COMPLETE_NOTIFY event
rather than gettimeofday(), which gives us the presentation time
instead of the time when SwapBuffers was called. Suggested by
Keith Packard. This relies on the fact that the X DRI3/Present
implementations use microseconds for UST.
v3: Properly ignore PresentCompleteKindMSCNotify; multiply in 64 bits
(caught by Keith Packard).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Keith Packard <keithp@keithp.com> [v3]
Reviewed-by: Marek Olšák <marek.olsak@amd.com> [v1]
These source files support actual geometry shaders, so using "gs" for
the name makes a lot of sense. We're going to be adding SIMD8 geometry
shader support as well, at which point "vec4_gs" will be a misnomer.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Matt Turner <mattst88@gmail.com>
Acked-by: Jason Ekstrand <jason.ekstrand@intel.com>
Acked-by: Iago Toral Quiroga <itoral@igalia.com>
The brw_gs.[ch] and brw_gs_emit.c source files contain code for
emulating fixed-function unit functionality (VF primitive decomposition
or SOL) using the GS unit. They do not contain code to support proper
geometry shaders.
We've taken to calling that code "ff_gs" (see brw_ff_gs_prog_key,
brw_ff_gs_prog_data, brw_context::ff_gs, brw_ff_gs_compile,
brw_ff_gs_prog). So it makes sense to make the filenames match.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Matt Turner <mattst88@gmail.com>
Acked-by: Jason Ekstrand <jason.ekstrand@intel.com>
Acked-by: Iago Toral Quiroga <itoral@igalia.com>
The GL functions and driver hooks use corresponding names---for example,
glMapBufferRange and Driver.MapBufferRange. But our implementation was
called "intel_bufferobj_map_range," which has the words "map" and
"buffer" swapped, as well as randomly adding "obj."
FlushMappedBufferRange was even trickier: it ordered the words
3, "obj", 1, 2, 4: intel_bufferobj_flush_mapped_range.
Even though the old names were consistent, I always had trouble
rearranging the jumble of words when searching for a function,
and it took a few tries to eventually land there.
The new names match the word order of GL and the driver hooks;
FlushMappedBufferRange is simply brw_flush_mapped_buffer_range.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
The OpenGL 4.0 core profile specification, section 2.17.3
Transform Feedback Draw Operations says:
"The error INVALID_VALUE is generated if <stream> is greater
than or equal to the value of MAX_VERTEX_STREAMS.
...
The error INVALID_OPERATION
is generated if EndTransformFeedback has never been called
while the object named by id was bound."
Fixes the piglit test:
ARB_transform_feedback3/arb_transform_feedback3-draw_using_invalid_stream_index
(with the test itself fixed to eliminate an unrelated failure)
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
When we're checking if the framebuffer is sRGB capable, call
is_format_supported() with the PIPE_BIND_DISPLAY_TARGET flag.
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
This reverts commit 20836c8185.
255 is a huge number. If you have a loop with 255 iterations, unrolling it
will exceed the SM3 instruction limit. Let's use the default again.
The comment about a SM3 limit doesn't make sense. For SM3, we generally
want 32 (default) or a lower number due to the SM3 instruction limit, which
is 512 instructions. For SM4, we can try higher numbers if needed, but
some shaders can end up being pretty huge and shader compilation can take
more time.
This fixes a shader compile failure on R500/SM3. Reported on IRC.
Cc: 10.2 10.3 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Caveat: Shaders using UBO/sampler indexing will
not be optimized by SB, due to SB not currently
supporting the necessary CF_INDEX_[01] index
registers.
Signed-off-by: Glenn Kennard <glenn.kennard@gmail.com>
The GL side of this extension just provides an accessor via glGetIntegerv for
the value of GL_CONTEXT_RELEASE_BEHAVIOR so it is trivial to implement. There
is a constant on the context for the value of the enum which is initialised to
GL_CONTEXT_RELEASE_BEHAVIOR_FLUSH. The extension is always enabled because it
doesn't need any driver interaction to retrieve the value.
If the value of the enum is anything but FLUSH then _mesa_make_current will
now refrain from calling _mesa_flush. This should only affect drivers that
explicitly change the enum to a non-default value.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
The main incentive to do this is to get the defines for the
GL_KHR_context_flush_control extension.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Different platforms require the offset to be in different units. However,
the generator fixes all of this up for us and only requires an offset in
bytes. Previously, we were getting this wrong all over the place. Some
computed/used it correctly as bytes while others treated the offset as
whole registers or computed it as bytes or bytes*2 in SIMD16 mode. This
commit cleans all this up and makes us properly treat it as bytes
everywhere.
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
I thought this would be a clever way to make spilling less expensive.
However, it appears that the oword read/write messages we are using for
spilling ignore the execution size and assume SIMD16 whenever working with
more than one register.
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Fanin (merge) nodes require it's srcs to be "adjacent" in consecutive
scalar registers. Keep track of instruction neighbors in copy-
propagation step and avoid eliminating mov's which would cause an
instruction to need multiple distinct left and/or right neighbors.
This lets us not fall on our face when we encounter things like:
1: MOV TEMP[2], IN[0].xyzw
2: TEX OUT[0].xy, TEMP[2], SAMP[0], SHADOW2D
3: MOV TEMP[2].xy, IN[0].yxzz
4: TEX OUT[0].zw, TEMP[2], SAMP[0], SHADOW2D
5: END
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Always insert extra mov's for the tex coord into the fanin. This
simplifies things a bit, and avoids a scenario where multiple sam
instructions can have mutually exclusive input's to it's fanin, for
example:
1: TEX OUT[0].xy, IN[0].xyxx, SAMP[0], 2D
2: TEX OUT[0].zw, IN[0].yxxx, SAMP[0], 2D
The CP pass can always remove the mov's that are not actually needed,
so better to start out with too many mov's in the front end, than not
enough.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
dscis -> noscis
dbypass -> nobypass
a bit more consistant w/ nobin, etc. And IMO a bit more sensible names.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Kills get added to the outputs list, to ensure they get scheduled. But
they aren't *really* outputs so skip them in the header comment block.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Adding this makes 'make check' catch failures introduced from
within ARB_clip_control.xml earlier.
Signed-off-by: Mathias Froehlich <Mathias.Froehlich@web.de>
In order to test compiler changes more easily, spit out the assembled
shader with some header information so that we can know about
inputs/outputs more easily.
See: git://people.freedesktop.org/~robclark/ir3test
In ir3test we have a big collection of tgsi shaders and reference
ir3_compiler outputs. When making compiler changes, regenerate the
compiler outputs and feed to ir3test to compare the new vs reference
shader.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
The last few dwords were skipped if the total number of dwords was not a
multiple of 4. Change the formatting for better readability.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
We only get here if the VS/GS compiled programs change, but we can even
skip it if the VS/GS size didn't change.
Affects cairo runtime on glamor by -1.26471% +/- 0.674335% (n=234)
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Just remove the parameter. Silences:
brw_program.c: In function 'brw_dump_ir':
brw_program.c:566:33: warning: unused parameter 'brw' [-Wunused-parameter]
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Originally I just fixed some unused parameter warnings in this
function. However, Ken pointed out:
"You could instead remove this driver hook. If the dd pointer is
NULL, arbprogram.c will return true. I think I'd prefer that."
Way, way back in time, I think _mesa_GetProgramivARB had the opposite
behavior. Given that it works the way it now works, I also prefer
removing the driver hook.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Just remove the parameter. Silences:
../../src/mesa/main/uniform_query.cpp:1062:1: warning: unused parameter 'ctx' [-Wunused-parameter]
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Silences:
../../src/mesa/main/shaderobj.c: In function '_mesa_init_shader_program':
../../src/mesa/main/shaderobj.c:239:46: warning: unused parameter 'ctx' [-Wunused-parameter]
For now, this adds a couple other unused parameter warnings, but future
patches will clean those up.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
_mesa_link_shader_program already calls _mesa_clear_shader_program_data
before calling link_shaders, so this is already done.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
_mesa_UseShaderProgramEXT, _mesa_ActiveProgramEXT, and
_mesa_CreateShaderProgramEXT were all removed when support for
GL_EXT_separate_shader_objects was removed.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Just remove the parameter. Silences:
../../src/mesa/main/ff_fragment_shader.cpp:668:1: warning: unused parameter 'p' [-Wunused-parameter]
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Previously, we were allowing the register allocation code to do the
computation for us in ra_set_finalize. However, the runtime for this
computation is O(c^4 * g) where c is the number of classes and g is the
number of GRF registers. However, these q-values are directly computable
based on the way we lay out our register classes so there is no need for
the aweful runtime algorithm.
We were doing ok until commit 7210583eb where we bumped the number of
register classes from 11 to 16. While startup times don't normally matter,
this caused piglit to take 4 times as long to run on Bay Trail. This patch
should make generating the ra_set much faster and melt the piglit run
times.
v2: Fixed a couple of bugs. I have now verified that the same q-values are
generated both ways.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
On older GENs in SIMD16 mode, we were accidentally building too much
interference into our register classes. Since everything is divided by 2,
the reigster allocator thinks we have 64 base registers instead of 128.
The actual GRF mapping still needs to be doubled, but as far as the ra_set
is concerned, we only have 64. We were accidentally adding way too much
interference.
Signed-off-by: Jason Ekstrand <jason.ekstrand@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
For GEN6 SIMD16 mode, we have to 2-align all the registers, so we only have
the even-numbered ones. This means that we have to divide the register
number by 2 when we precolor. This wasn't a problem before because we were
setting up the interference between ra_node registers wrong. This will be
fixed in the next commit.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
This shouldn't be a functional change since reg_belongs_to_class is just a
wrapper around BITSET_TEST. It just makes the code a little easier to
read.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Gallium should be prepared fine for ARB_clip_control.
So enable this and mention it in the release notes.
v2:
Only enable for drivers announcing the freshly introduced
PIPE_CAP_CLIP_HALFZ capability.
v3:
Use extension enable infrastructure to connect PIPE_CAP_CLIP_HALFZ
with ARB_clip_control.
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Mathias Froehlich <Mathias.Froehlich@web.de>
In preparation of ARB_clip_control. Let the driver decide if
it supports pipe_rasterizer_state::clip_halfz being set to true.
v3:
Initially enable on ilo.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Mathias Froehlich <Mathias.Froehlich@web.de
Restore clip control to the default state if MESA_META_VIEWPORT
or MESA_META_DEPTH_TEST is requested.
v3:
Handle clip control state with MESA_META_TRANSFORM.
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Mathias Froehlich <Mathias.Froehlich@web.de>
Implement the mesa parts of ARB_clip_control.
So far no driver enables this.
v3:
Restrict getting clip control state to the availability
of ARB_clip_control.
Move to transformation state.
Handle clip control state with the GL_TRANSFORM_BIT.
Move _FrontBit update into state.c.
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Mathias Froehlich <Mathias.Froehlich@web.de>
This is for preparation of ARB_clip_control.
v3:
Add comments.
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Mathias Froehlich <Mathias.Froehlich@web.de>
This allows vc4_opt_cse.c to CSE-away operations involving the same
uniform values.
total instructions in shared programs: 37341 -> 36906 (-1.16%)
instructions in affected programs: 10233 -> 9798 (-4.25%)
total uniforms in shared programs: 10523 -> 10320 (-1.93%)
uniforms in affected programs: 2467 -> 2264 (-8.23%)
This saves a bunch of extra flushes when texsubimaging a whole texture
that's been used for rendering, or subdataing a whole BO. In particular,
this massively reduces the runtime of piglit texture-packed-formats (when
the probes have been moved out of the inner loop).
I'm going to be using VC4_DEBUG=shaderdb,norast to do shaderdb stats, but
when debugging regressions, I want to match shaderdb output to shader
disassembly.
fd_bo_cpu_prep() doesn't realize the bo is already referenced in
unflushed cmdstream. It could be made to do so (but would have to be
implemented twice, ie. both for msm and kgsl). But we still can't do
the expected thing if the caller isn't using _NOSYNC. Because of the
way the tiling works, we need to build quite a bit of cmdstream at flush
time, which is not possible to do at the libdrm level.
So rather than trying to make fd_bo_cpu_prep() smarter than it can
possibly be, just *always* discard and reallocate if the
PIPE_TRANSFER_DISCARD_WHOLE_RESOURCE flag is set.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
- Use macro for munmap under Android - the STATIC_ASSERT uses
a off_t which is not used under Android for mmap. As loff_t size
does not vary as does off_t just ignore the assert.
- Wrap the long lines to improve readability.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
The setMCJITMemoryManager method doesn't exist in LLVM 3.3.
I thought I had tested the latest version of my earlier change with LLVM
3.3, but it looks I missed it.
Trivial.
JITMemoryManager was removed in LLVM 3.6, and replaced by its base class
RTDyldMemoryManager.
This change fixes our JIT memory managers specializations to derive from
RTDyldMemoryManager in LLVM 3.6 instead of JITMemoryManager.
This enables llvmpipe to run with LLVM 3.6.
However, lp_free_generated_code is basically a no-op because there are
not enough hook points in RTDyldMemoryManager to track and free the code
of a module. In other words, with MCJIT, code once created, stays
forever allocated until process destruction. This is not speicfic to
LLVM 3.6 -- it will happen whenever MCJIT is used regardless of version.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
We'll need to update gallivm for the interface changes in LLVM 3.6, and
the fewer the number of older LLVM versions we support the less hairy that
will be.
As consequence HAVE_AVX define can disappear. (Note HAVE_AVX meant
whether LLVM version supports AVX or not. Runtime support for AVX is
always checked and enforced independently.)
Verified llvmpipe builds and runs with with LLVM 3.3, 3.4, and 3.5.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
The functionality exposed by those extensions does not appear in ES3
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Matt Turner <mattst88@gmail.com>
If the user calls util_blitter_cache_all_shaders() set a flag and assert
that we never try to create any new fragment shaders after that point.
If the assertions fails, it means we missed generating some shader in
util_blitter_cache_all_shaders().
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Previously, we generated an extra CMP instruction:
cmp.ge.f0(8) g6<1>D g1<0,4,1>F 0F
cmp.nz.f0(8) null g6<4,4,1>D 0D
(+f0) sel(8) g5<1>F g1.4<0,4,1>F g2<0,4,1>F
The first operand is always a boolean, and we want to predicate the SEL
on that. Rather than producing a boolean value and comparing it against
zero, we can just produce a condition code in the flag register.
Now we generate:
cmp.ge.f0(8) null g1<0,4,1>F 0F
(+f0) sel(8) g5<1>F g1.4<0,4,1>F g2<0,4,1>F
No difference in shader-db.
v2: Remember to delete the old code (thanks Matt).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Using dst_reg(this, ir->type) automatically sets the writemask to the
proper size for the type; src_reg(dst_reg) preserves that. This should
be equivalent, but less code.
Note that src_reg(dst_reg) either uses SWIZZLE_XXXX or SWIZZLE_XYZW, so
the old code did need the manual writemask adjustment, since it
constructed the registers the other way around.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Nothing uses the vector_elements temporary variable.
Setting this->result.file is dead because we overwrite this->result a
few lines later.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Previously, we generated an extra CMP instruction:
cmp.ge.f0(8) g4<1>D g2<0,1,0>F 0F
cmp.nz.f0(8) null g4<8,8,1>D 0D
(+f0) sel(8) g120<1>F g2.4<0,1,0>F g3<0,1,0>F
The first operand is always a boolean, and we want to predicate the SEL
on that. Rather than producing a boolean value and comparing it against
zero, we can just produce a condition code in the flag register.
Now we generate:
cmp.ge.f0(8) null g2<0,1,0>F 0F
(+f0) sel(8) g124<1>F g2.4<0,1,0>F g3<0,1,0>F
total instructions in shared programs: 5473459 -> 5473253 (-0.00%)
instructions in affected programs: 6219 -> 6013 (-3.31%)
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
A while back, Matt made the uniform upload functions simply upload
ctx->Const.UniformBooleanTrue for boolean values instead of 0/1, which
removed the need to convert it later. We also set UniformBooleanTrue to
1.0f for drivers which want to treat booleans as 0.0/1.0f.
Nothing ever sets these, so they are dead.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
We don't have a scissor enable bit in hw, so when a raster state change
results in scissor enable bit changing, we need to also mark scissor
state as dirty.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
The optimization of avoiding restore (mem2gmem) if there was a clear
falls down a bit if you don't have a fullscreen scissor. We need to
make the decision logic a bit more clever to keep track of *what* was
cleared, so that we can (a) completely skip mem2gmem if entire buffer
was cleared, or (b) skip mem2gmem on a per-tile basis for tiles that
were completely cleared.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
DataLayoutPass was added in LLVM 3.5 r202168, commit
57edc9d4ff1648568a5dd7e9958649065b260dca "Make DataLayout a plain
object, not a pass.".
This patch fixes this build error with LLVM 3.4.
CXX llvm/libclllvm_la-invocation.lo
llvm/invocation.cpp: In function 'void {anonymous}::optimize(llvm::Module*, unsigned int, const std::vector<llvm::Function*>&)':
llvm/invocation.cpp:324:18: error: expected type-specifier
PM.add(new llvm::DataLayoutPass(mod));
^
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=85189
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
The values are hardcoded in the LLVM backend, but the TGSI definitions are
going to be changed with tessellation, e.g. TGSI_PROCESSOR_COMPUTE will be
increased by 2.
We'll use VS for LS and HS, because there's nothing special about them
from the LLVM backend point of view, even though the hardware side is
different. We do the same for ES.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
I'll need indexed loads without the meta data flag for tessellation later.
Also rename load_const to buffer_load_const to distinguish it from indexed
const loads.
v2: add comments
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
We must convert it to boolean from the DX9 float encoding that Gallium
specifies.
Later, we should probably define that FACE should be 0 or ~0 if native
integers are supported.
Cc: 10.2 10.3 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
With 5 shader stages and various combinations of enabled and disabled shaders,
the maximum number of outputs in one shader doesn't have to be equal to
the maximum number of inputs in the following shader.
v2: return 32 for softpipe and llvmpipe
If the writemask doesn't compress, then we want to put in the uncompressed
writemask, not the compressed writemask failure value (all-on).
Fixes glean's stencil2 and fbo-clear-formats on stencil.
It seems like the hardware is unhappy if we execute a kill instruction
prior to last input (ei). Probably the shader thread stops executing
and the end-input flag is never set.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
If app only updates (for example) vertex uniforms, it would be nice to
only re-emit those and not also frag uniforms. Means we need to mark
the first frag shader const buffer dirty after a clear.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
The get_variable_being_redeclared() function can free the 'var' argument.
Thereafter, we cannot assume that 'var' is a valid pointer. This patch
replaces 'var->name' with 'earlier->name' in two places and calls
is_gl_identifier(var->name) before 'var' might get freed.
This fixes several piglit GLSL crashes, including:
spec/glsl-1.50/execution/geometry/clip-distance-in-param
spec/glsl-1.50/execution/geometry/clip-distance-bulk-copy
spec/glsl-1.50/compiler/gs-redeclares-pervertex-out-before-global-redeclaration.geom
I'm not sure why these were not spotted sooner.
A similar bug was previously fixed by f9cecca7a.
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Patch fixes 'glsl-2types-of-textures-on-same-unit' in WebGL conformance
test suite. No Piglit regressions, fixes gl-2.0-active-sampler-conflict.
To avoid adding potentially heavy check during draw (valid_to_render),
check is done during uniform updates by inspecting TexturesUsed mask.
A new boolean variable is introduced to cache validation state.
v2: take into account case where 2 uniforms use same unit (curro)
also do the check only when SSO is not in use, SSO has own
path for sampler validation.
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Patch removes old variable based logic for handling a break inside
switch. Switch is put inside a loop so that existing infrastructure
for loop flow control can be used for the switch, now also dead code
elimination works properly.
Possible 'continue' call inside a switch needs now special handling
which is taken care of by detecting continue, breaking out and calling
continue for the outside loop.
v2: remove one unnecessary ir_expression (Curro)
Fixes following Piglit tests:
fs-exec-after-break.shader_test
fs-conditional-break.shader_test
No Piglit or es3conform regressions.
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
GLES2 doesn't have GL_TEXTURE_BASE_LEVEL, so the hardware doesn't. Fixes
piglit levelclamp, tex-miplevel-selection, and texture-storage/2D mipmap
rendering.
Suppresses a bunch of warning noise about sample_map possibly being used
uninitialized.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Before, we used the a signed d-word for booleans and the immedates we
emitted varried between signed and unsigned. This commit changes the type
to unsigned (I think that makes more sense) and makes immediates more
consistent. This allows copy propagation to work better cleans up some
instructions.
total instructions in shared programs: 5473519 -> 5465864 (-0.14%)
instructions in affected programs: 432849 -> 425194 (-1.77%)
GAINED: 27
LOST: 0
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
gl_SampleID is a built-in variable that always is of type "int".
Suggested by Connor Abbott.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Notably this included the EOF flag (the other bits are the full buffer
dump selection, but we don't do full dumps), which caused the kernel
checking for frame completion to trigger.
The other driver does this manually before calling into each tile, but we
can just let it get binned into the tiles (saving repeated kernel
validation on the packet).
Fixes simulator assertion failures on polygon-mode and non-auto texwrap.
We don't need to emit all of our current state at the end of each bin
list. We're going to be smashing it all at the start of the next tile's
bin list, anyway.
The trick is to generate a unique buffer usage value for each possible
combination of domains and flags, with only one bit set each for the
domains and flags. This ensures pb_check_usage() only returns TRUE when
the domains and flags the cached buffer was created for exactly match
the requested ones.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
There are two debug variables:
CLOVER_DEBUG which you can set to any combination of llvm,clc,asm
(separated by commas) to dump llvm IR, OpenCL C, and native assembly.
CLOVER_DEBUG_FILE which you can set to a file name for dumping output
instead of stderr. If you set this variable, the output will be split
into three separate files with different suffixes: .cl for OpenCL C,
.ll for LLVM IR, and .asm for native assembly. Note that when data
is written, it is always appended to the files.
v2:
- Code cleanups
- Add CLOVER_DEBUG_FILE environment variable for dumping to a file.
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
v2:
- Split build_module_native() into three separate functions.
- Code cleanups.
v3:
- More cleanups.
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Drivers can return this value for PIPE_COMPUTE_CAP_IR_TARGET
if they want clover to give them native object code.
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
NewBufferObject took a "target" parameter, which it blindly passed to
_mesa_initialize_buffer_object(), which ignored it.
Not much point in passing it around.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
This is used to implement GLSL's atomicCounter() intrinsic. Previously
it *worked*, but the disassembly was bogus.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
This would have *almost never* actually been an issue, since other state
tends to get flagged at the same time as new ABOs -- but still bogus.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
This didn't make any sense, but papered over the missing TexBO flagging
we've just fixed, in a bunch of cases.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
When a buffer object is bound to one of the indexed uniform buffer
binding points, assume that from that point on it may be used as
a uniform buffer.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
In the drivers, we occasionally want to reallocate the backing
store for a buffer object; often to avoid waiting for the GPU
to be finished with the previous contents.
At the point that happens, we don't have a good way of determining
where else the buffer object may be bound, and so no good way of
determining which dirty flags need to be raised -- it's fairly
expensive to go looking at all the possible binding points.
Until now, we've considered any BO to be possibly bound as a UBO or
TexBO, and flagged all that state to be reemitted.
Instead, remember what kinds of binding point this buffer has ever
been used with, so that the drivers can flag only what they need.
I don't expect these bits to ever be reset, but that doesn't matter
for reasonable apps.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
The kernel files are built into a separate static library and
all the functions that require it are already wrapped in ifdef
USE_VC4_SIMULATOR. Don't forget the header file :)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Now that we've made all the texture emit code mostly independent of GLSL
IR, this isn't necessary any more.
Signed-off-by: Connor Abbott <connor.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Before, we had 3 different emit functions for various different gen's,
as well as some ancilliary work that was the same across all gen's which
was either contained in functions or duplicated across the GLSL IR and
Mesa IR backends. Now, we have a single method, emit_texture(), that
takes all the information needed to make a texture instruction and
handles all the setup, and all we have to do to emit a texture
instruction while converting from GLSL IR, Mesa IR, or any new backend
is to extract the information emit_texture() needs and then call it.
v2: Significant rebasing (by Ken).
Signed-off-by: Connor Abbott <connor.abbott@intel.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
This happened to work before, but it would convert the output to a float
and then back to an integer which seems bad.
Signed-off-by: Connor Abbott <connor.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
This drops a dependency on ir_texture objects.
v2 (Ken): Rename lod_components to grad_components, as it only has a
meaningful value for ir_txd. We could set it to 1 for TXL,
but there's no real need.
Signed-off-by: Connor Abbott <connor.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Our new IR won't have ir_texture objects, but using glsl_type is fine.
v2 (Ken): Drop redundant ir->coordinate NULL check; rebase.
Signed-off-by: Connor Abbott <connor.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
This is slightly clearer. Based on a patch by Connor Abbott.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
This moves the handling of non-constant texel offset subexpression trees
to the place where we visit other such subtrees. It also removes some
uses of ir->offset in emit_texture_gen7, which will be useful when we
write the backend for our new upcoming IR.
Based on a patch by Connor Abbott.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
brw_lower_unnormalized_offset sets ir->offset to NULL if it applies the
texelFetchOffset workarounds, so there's no need to special case it
here---there won't be an offset for ir_txf.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Eric's original code to work around TXF offset bugs contained a comment
explaining the problem, which was lost when Chris generalized it to an
IR transformation (in commit 598ca510b8).
This commit adds the original comment to the newer code.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Because we reuse various bits of emit code (for state/vertex/prog/etc)
for both regular draws and internal draws (gmem<->mem, clear, etc), the
number of parameters getting passed around has been growing. Refactor
to group these into fd3_emit. This simplifies fxn signatures, avoids
passing around shader key on the stack, etc. It also gives us a nice
place to cache shader-variant lookup to avoid looking up shader variants
multiple times per draw (without having to *also* pass them around as
fxn args everywhere).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Get rid of fd3_vertex_buf and use fd_vertex_state directly for all
draws. Removes a tiny bit of CPU overhead for munging around the vertex
state every time it is emitted, but more importantly it cleans things up
for later optimizations, so the emit paths don't have to special case
internal draws (gmem<->mem, clears, etc) with regular draws.
Instead of constructing fd3_vertex_buf array each time for internal
draws, and context init time pre-create solid_vbuf_state and
blit_vbuf_state.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Due to the implicit move-from-GRF, unary math looks a lot like the Gen6+
math instruction: it's a single instruction (SEND) with a GRF source.
The difference is that it also implicitly clobbers a message register.
The only visible effect is that CSE will remove the MRF-clobbering from
later math operations. This should be fine; compute_to_mrf and
remove_redundant_mrf_writes don't look at the values populated by
implied writes, so they can't rely on those values being present.
Less interference may actually help those passes make more progress.
Binary math is still problematic, since it involves a separate MOV
instruction to load the second operand. We continue disallowing CSE for
binary math operations.
total instructions in shared programs: 3340303 -> 3340100 (-0.01%)
instructions in affected programs: 26927 -> 26724 (-0.75%)
Nothing hurt, gained, or lost. ~6% reduction on a few shaders.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Otherwise the caching buffer manager may return a buffer which was created
with a different set of flags, which can cause trouble.
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Previously, when no pkg-config was available for
libexpat we would just add the needed linking
flags without any extra check.
Now, we check that the library and the headers are
also installed in the building environment.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Now that the freedreno_lowering code is moved to tgsi_lowering, remove
our private copy and switch over to using the common version.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Originally the variables were set only once via the ?= operator but
that causes issues when doing incremental builds. They appear to be
undefined and missing from the dependency list despite their addition
to LIBADD.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84807
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
The texturing hardware takes the POT level 0 width/height and minifies
those. This is different from what we were doing, for example, for
273-wide's level 5: POT(273>>5) == 8, while POT(273)>>5 == 16.
Fixes piglit-depthstencil-render-miplevels 273.
This patch fixes this build error on DragonFly BSD.
CC os/os_misc.lo
os/os_misc.c: In function 'os_get_total_physical_memory':
os/os_misc.c:132:2: error: #error Unsupported *BSD
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
The current implementation of glxUseXFont requires creating
a temporary pixmap and graphics context, which requires a real
old-school X11 Window, not a glxDrawable. This patch changes
things so that glxUseXFont will also accept a glxWindow or
glxPixmap, and lookup the underlying X11 Drawable. Without
this patch glxUseXFont generates a giant stream of Xerrors
about bad drawables and bad graphics contexts.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=54372
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Brian Paul <brianp@vmware.com>
It does not look like an issue now but it is good to be future proof. Spotted
by Courtney Goeltzenleuchter.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
This gets them to pass glsl-sin/cos. There was an obvious problem that I
was using the FRC code on the scaled input value, which means that we had
a range in [0, 1], while our taylor is most accurate across [-0.5, 0.5].
We can just slide things over, but that means flipping the sign of the
coefficients. After that, it was just a matter of stuffing more
coefficients in.
There's no reason to stall on pwrite - the CPU always appends to the
buffer and never modifies existing contents, and the GPU never writes
it. Further, the CPU always appends new data before submitting a batch
that requires it.
This code predates the unsynchronized mapping feature, so we simply
didn't have the option when it was written.
Ideally, we would do this for non-LLC platforms too, but unsynchronized
mapping support only exists for LLC systems.
Saves a bunch of stall avoidance copies when uploading shaders.
v2: Rebase on changes to previous patch.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net> [v1]
We don't really want unnecessary buffer copying, so it'd be nice to know
when it's happening.
v2: Drop stall warnings when doing a read-only CPU mapping of the cache
BO. The GPU also uses it in a read-only fashion, so there won't be
any stalls, even though the buffer is busy. (Thanks to Chris Wilson
for catching this mistake.)
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net> [v1]
This is easy: we just need to use brw_map_bo instead of mapping it
directly.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
If the VS doesn't output a value that the FS needs, we still need to read
the right contents for the remaining FS inputs, by emitting padding. And
if the VS outputs something the FS doesn't need, we shouldn't put it in
the VPM at all (so the code producing it can get DCEed).
Fixes 77 piglit tests.
Also fixes two sided lighting which was broken at least
on pre-evergreen by commit b1eb00.
Signed-off-by: Glenn Kennard <glenn.kennard@gmail.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
This is the last use tgsi_parse_token in radeonsi.
It looks ugly because the code was re-indented, but there is really no change
in behavior.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
They were reinventing tgsi_shader_info. They are unused now.
radeon_llvm_context::load_input can be NULL if input fetching is implemented
in some other way.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
No code in Mesa sets the usage mask to any other value.
The final mask is AND'ed with enable bits from the rasterizer state anyway.
If somebody implements setting usage masks in st/mesa, we can use
tgsi_shader_info to get it more easily.
This is a prerequisite for the following commit.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
[ Francisco Jerez: Split off from a larger patch, and take a slightly
different approach for passing the implicit arguments around. ]
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Our current atan()-approximation is pretty inaccurate at 1.0, so
let's try to improve the situation by doing a direct approximation
without going through atan.
This new implementation uses an 11th degree polynomial to approximate
atan in the [-1..1] range, and the following identitiy to reduce the
entire range to [-1..1]:
atan(x) = 0.5 * pi * sign(x) - atan(1.0 / x)
This range-reduction idea is taken from the paper "Fast computation
of Arctangent Functions for Embedded Applications: A Comparative
Analysis" (Ukil et al. 2011).
The polynomial that approximates atan(x) is:
x * 0.9999793128310355 - x^3 * 0.3326756418091246 +
x^5 * 0.1938924977115610 - x^7 * 0.1173503194786851 +
x^9 * 0.0536813784310406 - x^11 * 0.0121323213173444
This polynomial was found with the following GNU Octave script:
x = linspace(0, 1);
y = atan(x);
n = [1, 3, 5, 7, 9, 11];
format long;
polyfitc(x, y, n)
The polyfitc function is not built-in, but too long to include here.
It can be downloaded from the following URL:
http://www.mathworks.com/matlabcentral/fileexchange/47851-constraint-polynomial-fit/content/polyfitc.m
This fixes the following piglit test:
shaders/glsl-const-folding-01
Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
When mapping the buffer a second time, we need to use the new pointer,
not the one from the previous mapping. Otherwise, we will most likely
crash.
Apparently, we've just been getting lucky and getting the same
bo->virtual pointer in both cases. libdrm probably has a hand in that.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Cc: mesa-stable@lists.freedesktop.org
Copy propagating these might result in reading the r4 after some other
instruction has written r4. Just prevent all copy propagation of this for
now.
Fixes bad rendering with upcoming indirect register access support, where
the copy propagation was consistently happening across another read.
Merging VS and CS into the same struct wasn't winning us anything except
for not allocating a separate BO (but if we want to pack programs into
BOs, we should pack not just those 2 programs together). What it was
getting us was a bunch of code duplication about hash table lookups and
propagating vc4_compile contents into a vc4_compiled_shader.
I was about to make the situation worse with indirect uniform buffer
access.
I wanted to make another set of texture uploads for handling reladdr
constants, and duplicating all the bitshifting looked like a terrible
idea. In the process, this fixes a swap of the s/t texture wrap modes.
Under the simulator, reading registers before writing them triggers an
assertion failure. c->undef gets treated as r0, which will usually be
written, but not if it's used in the first instruction. We should
definitely not be aborting in this case, and return some sort of undefined
value instead.
Fixes glsl-user-varying-ff.
The border color is only needed when using the GL_CLAMP_TO_BORDER or
(deprecated) GL_CLAMP wrap modes; all others ignore it, including the
common GL_CLAMP_TO_EDGE and GL_REPEAT wrap modes.
In those cases, we can skip uploading it entirely, saving a bit of space
in the batchbuffer. Instead, we just point it at the start of the
batch (offset 0); we have to program something, and that address is safe
to read.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Write-back caching cannot be used for buffers being scanned out by the
display engine; surfaces used for scan-out must be write-through or
uncached. I originally chose WT for render targets because it works in
all cases. However, we really want to use write-back caching where
possible, as it is more efficient.
Most renderbuffers are not used for scanout - off-screen FBOs certainly
are fine, and non-pageflipped backbuffers should be fine as well. So
in most cases WB will work. However, we don't know what will be used
for scan-out, so we instead simply use the PTE value specified by the
kernel, as it knows these things.
This matches our MOCS choice on Haswell.
Fixes performance regressions since commit ee4484be3d
in a microbenchmark (spotted by Eero Tamminen). Improves performance
in GLBenchmark 2.7/EgyptHD by 7.44362% +/- 0.496939% (n=55) on a
Broadwell GT2. Improves performance in a bunch of other microbenchmarks
by ~15% or so.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reported-by: Eero Tamminen <eero.t.tamminen@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Cc: mesa-stable@lists.freedesktop.org
Like BDW_MOCS_WB and BDW_MOCS_WT, this specifies that we want to use all
three caches (L3, LLC, and eLLC where available), but leaves the LLC
caching mode up to the kernel's page table entry.
This allows the kernel to pick WB/WT/UC based on whether it's using a
buffer for scanout.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Cc: mesa-stable@lists.freedesktop.org
These days, most driver debug output happens via stderr, not stdout.
Some applications (such as Xephyr) also appear to close stdout which
makes these messages go nowhere.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
The non-base NPOT levels are stored as POT-aligned images. We get that
POT alignment by minifying the POT-aligned base level.
This means that level strides are also POT aligned, so we have to tell the
rendering mode config that our resource is larger than the actual
requested area.
Fixes the fbo-generatemipmap-formats NPOT cases. Regresses
depthstencil-render-miplevels 273 * -- the texture presentation now works
(where it was completely broken before), it looks like there's some
overflow of image bounds happening at the lower miplevels.
It's fairly easy, thanks to Rob Clark's lowering code. Fixes
two-sided-lighting and 4 vertex-program-two-side testcases, while
regressing 8 testcases that involve enabling two-sided color while only
initializing one of the two colors in the VS. If you're enabling two
sided color, it's of course expected that you really do set up both
colors, so this is still an improvement (and when we set up a linker for
TGSI, we'll hopefully fix those 8 fails).
Lots of drivers need to transform the weird instructions in TGSI into
reasonable scalar ops, and this code can make those translations
canonical.
Acked-by: Rob Clark <robclark@freedesktop.org>
The simulator assertion fails if you have a write to a reg and then a read
(for example, in the NOP side of an instruction), even if the read isn't
used for anything. By setting unused raddrs to NOP, we avoid the problem
(since only the phsyical registers are tracked).
Original patch by Petri Latvala <petri.latvala@intel.com>:
Add an optimization pass that drops min/max expression operands that
can be proven to not contribute to the final result. The algorithm is
similar to alpha-beta pruning on a minmax search, from the field of
AI.
This optimization pass can optimize min/max expressions where operands
are min/max expressions. Such code can appear in shaders by itself, or
as the result of clamp() or AMD_shader_trinary_minmax functions.
This optimization pass improves the generated code for piglit's
AMD_shader_trinary_minmax tests as follows:
total instructions in shared programs: 75 -> 67 (-10.67%)
instructions in affected programs: 60 -> 52 (-13.33%)
GAINED: 0
LOST: 0
All tests (max3, min3, mid3) improved.
A full shader-db run:
total instructions in shared programs: 4293603 -> 4293575 (-0.00%)
instructions in affected programs: 1188 -> 1160 (-2.36%)
GAINED: 0
LOST: 0
Improvements happen in Guacamelee and Serious Sam 3. One shader from
Dungeon Defenders is hurt by shader-db metrics (26 -> 28), because of
dropping of a (constant float (0.00000)) operand, which was
compiled to a saturate modifier.
Version 2 by Iago Toral Quiroga <itoral@igalia.com>:
Changes from review feedback:
- Squashed various cosmetic changes sent by Matt Turner.
- Make less_all_components return an enum rather than setting a class member.
(Suggested by Mat Turner). Also, renamed it to compare_components.
- Make less_all_components, smaller_constant and larger_constant static.
(Suggested by Mat Turner)
- Change mixmax_range to call its limits "low" and "high" instead of
"range[0]" and "range[1]". (Suggested by Connor Abbot).
- Use ir_builder swizzle helpers in swizzle_if_required(). (Suggested by
Connor Abbot).
- Make the logic more clearer by rearrenging the code and commenting.
(Suggested by Connor Abbot).
- Added comment to explain why we need to recurse twice. (Suggested by
Connor Abbot).
- If we cannot prune an expression, do not return early. Instead, attempt
to prune its children. (Suggested by Connor Abbot).
Other changes:
- Instead of having a global "valid" visitor member, let the various functions
that can determine this status return a boolean and check for its value
to decide what to do in each case. This is more flexible and allows to
recurse into children of parents that could not be prunned due to invalid
ranges (so related to the last bullet in the review feedback).
- Make sure we always check if a range is valid before working with it. Since
any use of get_range, combine_range or range_intersection can invalidate
a range we should check for this situation every time we use any of these
functions.
Version 3 by Iago Toral Quiroga <itoral@igalia.com>:
Changes from review feedback:
- Now we can make get_range, combine_range and range_intersection static too
(suggested by Connor Abbot).
- Do not return NULL when looking for the larger or greater constant into
mixed vector constants. Instead, produce a new constant by doing a
component-wise minmax. With this we can also remove of the validations when
we call into these functions (suggested by Connor Abbot).
- Add a comment explaining the meaning of the baserange argument in
prune_expression (suggested by Connor Abbot).
Other changes:
- Eliminate minmax expressions operating on constant vectors with mixed values
by resolving them.
No piglit regressions observed with Version 3.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76861
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Patch fixes following test case from 'shaders-with-varyings' WebGL
conformance suite: "vertex shader with unused varying and fragment
shader with used varying must succeed"
v2: emit still a warning if the condition happens (Ian)
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
When a shader needs N surfaces, we should upload N surfaces and not depend on
how many are bound. This commit is larger than it should be because we did
not export how many surfaces a surface uses before.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
It only needs the constant buffer with clip planes and read-write resources
for the GS->VS ring and streamout. That's 2 pointers.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
This is a wrong place to flush caches to say the least.
I don't think we need to flush the instruction caches if we don't patch
shaders with DMA.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
st/mesa has the same flag in its shader key, we don't need to do it
in the driver anymore.
Instead, use TGSI_INTERPOLATE_LOC_SAMPLE, which is what st/mesa sets.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
The first compiled shader is sometimes useless, because the key doesn't match
the key for the draw call where it's used.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Fixes a few issues, including a potential empty-IB (which triggers gpu
hangs in piglit occlusion_query_meta_no_fragments)
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Possibly we should map the front color to black (zeroes). But not sure
there is a way to do that without generating a shader variant.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Shaders like:
FRAG
PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
DCL IN[0], GENERIC[0], PERSPECTIVE
DCL OUT[0], COLOR
DCL SAMP[0]
DCL TEMP[0], LOCAL
IMM[0] FLT32 { 0.0000, 1.0000, 0.0000, 0.0000}
0: TEX TEMP[0], IN[0].xyyy, SAMP[0], 2D
1: MOV OUT[0], IMM[0].xyxx
2: END
cause unhappyness. They have an IN[], but once this is compiled the
useless TEX instruction goes away. Leaving a varying that is never
fetched, which makes the hw unhappy.
In the process fix a signed vs unsigned compare. If the vertex shader
has max_reg=-1, MAX2() vs an unsigned would not give the desired result.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
On Windows, the Piglit primitive-restart test was failing a
glGetError()==0 assertion when it was run w/out any command line
arguments. Piglit's all.py script only runs primitive-restart
with arguments so this case isn't normally hit during a full
piglit run.
The basic problem is Microsoft's opengl32.dll calls glFlush
from wglGetProcAddress() and Piglit uses wglGetProcAddress() to
resolve glPrimitiveRestartNV() which is called inside glBegin/End.
See comments in the code for more info.
Plus, improve the comments for _mesa_alloc_dispatch_table().
Cc: <mesa-stable@lists.freedesktop.org>
Acked-by: Sinclair Yeh <syeh@vmware.com>
This isn't perfect -- the filtering is happening on the srgb values, and
we're decoding afterwards, which is not what you want. I think that's the
cause of some additional texwrap(GL_CLAMP, LINEAR) failures, though many
other texwrap tests on srgb start to pass since unfiltered values come out
correct.
Since the RA has to be done s.t. each one gets its own (adjacent)
register, it would complicate matters if instructions were allowed to be
repeated. This enables copy-propagation use in situations where
previously that might have happened.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
While running piglit in virgl, I hit an assert in intel driver.
"qemu-system-x86_64: intel_tex.c:219: intel_map_texture_image: Assertion `tex_image->TexObject->Target != 0x8C18 || h == 1' failed."
Thanks to Eric and Ken for pointing me in the right direction,
Fix the get_tex_depth to do the same fixup as get_tex_rgba does
for 1D array textures.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
With current makefiles the build fails because source and build paths
are generated incorrectly. With Android build system the top_srcdir and
top_builddir variables are undefined and all paths are relative to where
Android.mk is located. This ends up with path likes
external/mesa/src/mesa/src/mesa/ for both source and build paths, which
are obviously wrong.
This patch fixes this by overriding resulting SRCDIR and BUILDDIR
variables with empty string, so that paths end up being relative to
Android.mk file again. Appending correct build path to generated files
is already done in Android.gen.mk.
Signed-off-by: Tomasz Figa <tomasz.figa@gmail.com>
CC: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This patch fixes Android build failures by including src/util directory
in compilation. Files inside of this directory are compiled into
libmesa_util static library and linked with resulting libGLES_mesa.
Signed-off-by: Tomasz Figa <tomasz.figa@gmail.com>
CC: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Before, we were hard-coding the base_mrf based on dispatch width not number
of registers spilled at a time. This caused us to emit instructions with a
base_mrf or 14 and a mlen of 3 so we used the magical non-existant m16
register. This fixes the problem.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Previously, we had a MAX_SAMPLER_MESSAGE_SIZE which we used instead.
However, some FB write messages can validly be longer than this so we need
something different. Since MAX_SAMPLER_MESSAGE_SIZE is validly useful on
its own, we leave it alone and add a new MAX_GRF_SIZE that's big enough for
FB writes.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84539
Reviewed-by: Matt Turner <mattst88@gmail.com>
This fixes a bug where 1-wide operations don't properly translate down to
1-wide instructions.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Above a certain limit use CACHE mode instead of BUFFER mode. This
should solve gpu hangs with large shader programs.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
The existing code was not setting several fields, most importantly the
target, which is required on nv50/nvc0.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Patch fixes failing test in WebGL conformance test
'point-no-attributes' when running Chrome on OpenGL ES.
(Shader program may draw points using constant data in shader.)
No Piglit regressions.
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
They're actually as documented in the HW specs and the GL mipmapping enums
order. Fixes fbo-generatemipmap-filtering , and some other tests where we
were off by a few bits due to unexpected linear filtering.
Extension enables doing a multisample buffer resolve and buffer
scaling using a single glBlitFrameBuffer() call. Currently, we
have this extension implemented in BLORP which is only used by
SNB and IVB. This patch implements the extension in meta path
which makes it available to Broadwell.
Implementation features:
- Supports scaled resolves of 2X, 4X and 8X multisample buffers.
- Avoids unnecessary shader compilations by storing the pre compiled
shaders for each supported sample count.
- Uses bilinear filtering for both GL_SCALED_RESOLVE_FASTEST_EXT and
GL_SCALED_RESOLVE_NICEST_EXT filter options. This is an allowed
behavior in the extension's spec.
- I tried doing bicubic filtering for GL_SCALED_RESOLVE_NICEST_EXT
filter. It made the edges in the image look little smoother but
the image gets blurred causing no overall quality improvement.
For now I have dropped the idea of doing different filtering for
nicest filter.
V2:
- Minor changes to simplify the fragment shader.
- Refactor the code to move i965 specific sample_map computation out
of Meta. We now use ctx->Const.SampleMap{2,4,8}x variables initialized
by the driver.
- Use a simple msaa resolve shader for scaled resolves with scaling
factor = 1.0.
V3:
- Make changes to create a string out of ctx->Const.SampleMap{2,4,8}x
variables and use it in fragment shader.
V4:
- Make changes to use uint8_t type ctx->Const.SampleMap{2,4,8}x
variables.
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
with values specific to Intel hardware.
V2: Define and use gen6_get_sample_map() function to initialize
the variables.
V3: Change the function name to gen6_set_sample_maps() and use
memcpy() to fill in the data.
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
SampleMap{2,4,8}x variables are used in later patches to implement
EXT_framebuffer_multisample_blit_scaled extension.
V2: Use integer array instead of a string.
Bump up the comment.
V3: Use uint8_t type array.
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
This patch implements functions for images support,
which basically supports copy data between video
surface and user buffers, in this case supports
SW decode, and other video output
v2: fix buffer size for odd-sized image case
expose I420 format as well
v3: fix YUV 4:2:2 format data buffer size
cleanup I420 format exposure
Signed-off-by: Leo Liu <leo.liu@amd.com>
This patch implements codec for mpeg2 h264 and vc1,
populates codec parameters and pass them to HW driver.
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Leo Liu <leo.liu@amd.com>
This patch implements context managements, relate it HW driver,
functions for video surface managements, and functions for
application data memory buffer managements.
implemented functions:
vlVa(Create|Destroy)Context
vlVa(Create|Destroy|Put)Surfaces
vlVa(Create|Destroy)Buffer
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Leo Liu <leo.liu@amd.com>
This patch is for application to query configuration,
such as profiles, entrypoints, and attributes
v2: fix missing profile with query
Signed-off-by: Michael Varga <michael.varga@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Leo Liu <leo.liu@amd.com>
This patch adds a skeleton VA-API state tracker,
which is filled with live in the subsequent patches.
v2: fixes in configure.ac and va state_tracker Makefile.am
v3: do not link against libva.
detect libva version, and correctly set driver entrypoint name.
rebase(cleanup) targets/va/Makefile.am
v4: cleanup va version auto detection
add back targets/va/va.sym
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Break out these functions so that they can be shared with a other
state trackers. They will be used in subsequent patches for the new
VA-API state tracker.
Signed-off-by: Leo Liu <leo.liu@amd.com>
Cuts the number of i965 color calculator viewport uploads by 100x
(11017983 -> 113385) in 'x11perf -gc' with Glamor in Xephyr.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
I believe when I wrote this code, gen6_sf_state used CACHE_NEW_VS_PROG,
which has since been replaced by BRW_NEW_VUE_MAP_GEOM_OUT. It's not
needed here anyway - only SBE needs it. Just a copy and paste mistake.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
This function flagged BRW_NEW_*_PROGRAM
When ctx->{Vertex,Geometry,Fragment}Program._Current changes, core Mesa
calls the BindProgram driver hook, which flagged BRW_NEW_*_PROGRAM.
However, brw_upload_state also checks for that changing, sets the same
flags, and also updates brw->fragment_program and so on. So, this looks
to be entirely redundant.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Now that the bitfield is a uint64_t, we should use 1ull. Currently, we
only have 32 entries, so 1 works fine, but it's not future-proof.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
This will keep INTEL_DEBUG=state working when we add BRW_NEW_* bits
beyond 1 << 31. We missed doing this when widening the driver flags
from uint32_t to uint64_t.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Fix build error introduced with commit
eedbce9c63.
lp_test_format.c: In function ‘test_format_unorm8’:
lp_test_format.c:226:4: error: too few arguments to function ‘gallivm_create’
gallivm = gallivm_create("test_module_unorm8");
^
In file included from ../../../../src/gallium/auxiliary/gallivm/lp_bld_format.h:38:0,
from lp_test_format.c:42:
../../../../src/gallium/auxiliary/gallivm/lp_bld_init.h:58:1: note: declared here
gallivm_create(const char *name, LLVMContextRef context);
^
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=84538
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Instead of just segfaulting in the driver when a buffer allocation fails,
report error messages indicating what went wrong so that we can debug things.
As a simple example, chromium wraps Mesa in a sandbox which doesn't allow
access to most syscalls, including the ability to create shared memory
segments for fences. Before, you'd get a simple segfault in mesa and your 3D
acceleration would fail. Now you get:
$ chromium --disable-gpu-blacklist
[10618:10643:0930/200525:ERROR:nss_util.cc(856)] After loading Root Certs, loaded==false: NSS error code: -8018
libGL: pci id for fd 12: 8086:0a16, driver i965
libGL: OpenDriver: trying /local-miki/src/mesa/mesa/lib/i965_dri.so
libGL: Can't open configuration file /home/keithp/.drirc: Operation not permitted.
libGL: Can't open configuration file /home/keithp/.drirc: Operation not permitted.
libGL error: DRI3 Fence object allocation failure Operation not permitted
[10618:10618:0930/200525:ERROR:command_buffer_proxy_impl.cc(153)] Could not send GpuCommandBufferMsg_Initialize.
[10618:10618:0930/200525:ERROR:webgraphicscontext3d_command_buffer_impl.cc(236)] CommandBufferProxy::Initialize failed.
[10618:10618:0930/200525:ERROR:webgraphicscontext3d_command_buffer_impl.cc(256)] Failed to initialize command buffer.
This made it pretty easy to diagnose the problem in the referenced bug report.
Bugzilla: https://code.google.com/p/chromium/issues/detail?id=415681
Signed-off-by: Keith Packard <keithp@keithp.com>
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Matt Turner <mattst88@gmail.com>
A driver which doesn't have async flip support will queue up flips without any
way to replace them afterwards. This means we've got a scanout buffer pinned
as soon as we schedule a flip and so we need another buffer to keep from
stalling.
When vblank_mode=0, if there are only three buffers we do:
current scanout buffer = 0 at MSC 0
Render frame 1 to buffer 1
PresentPixmap for buffer 1 at MSC 1
This is sitting down in the kernel waiting for vblank to
become the next scanout buffer
Render frame 2 to buffer 2
PresentPixmap for buffer 2 at MSC 1
This cannot be displayed at MSC 1 because the
kernel doesn't have any way to replace buffer 1 as the pending
scanout buffer. So, best case this will get displayed at MSC 2.
Now we block after this, waiting for one of the three buffers to become idle.
We can't use buffer 0 because it is the scanout buffer. We can't use buffer 1
because it's sitting in the kernel waiting to become the next scanout buffer
and we can't use buffer 2 because that's the most recent frame which will
become the next scanout buffer if the application doesn't manage to generate
another complete frame by MSC 2.
With four buffers, we get:
current scanout buffer = 0 at MSC 0
Render frame 1 to buffer 1
PresentPixmap for buffer 1 at MSC 1
This is sitting down in the kernel waiting for vblank to
become the next scanout buffer
Render frame 2 to buffer 2
PresentPixmap for buffer 2 at MSC 1
This cannot be displayed at MSC 1 because the
kernel doesn't have any way to replace buffer 1 as the pending
scanout buffer. So, best case this will get displayed at MSC
2. The X server will queue this swap until buffer 1 becomes
the scanout buffer.
Render frame 3 to buffer 3
PresentPixmap for buffer 3 at MSC 1
As soon as the X server sees this, it will replace the pending
buffer 2 swap with this swap and release buffer 2 back to the
application
Render frame 4 to buffer 2
PresentPixmap for buffer 2 at MSC 1
Now we're in a steady state, flipping between buffer 2 and 3
waiting for one of them to be queued to the kernel.
...
current scanout buffer = 1 at MSC 1
Now buffer 0 is free and (e.g.) buffer 2 is queued in
the kernel to be the scanout buffer at MSC 2
Render frames, flipping between buffer 0 and 3
When the system can replace a queued buffer, and we update Present to take
advantage of that, we can use three buffers and get:
current scanout buffer = 0 at MSC 0
Render frame 1 to buffer 1
PresentPixmap for buffer 1 at MSC 1
This is sitting waiting for vblank to become the next scanout
buffer
Render frame 2 to buffer 2
PresentPixmap for buffer 2 at MSC 1
Queue this for display at MSC 1
1. There are three possible results:
1) We're still before MSC 1. Buffer 1 is released,
buffer 2 is queued waiting for MSC 1.
2) We're now after MSC 1. Buffer 0 was released at MSC 1.
Buffer 1 is the current scanout buffer.
a) If the user asked for a tearing update, we swap
scanout from buffer 1 to buffer 2 and release buffer
1.
b) If the user asked for non-tearing update, we
queue buffer 2 for the MSC 2.
In all three cases, we have a buffer released (call it 'n'),
ready to receive the next frame.
Render frame 3 to buffer n
PresentPixmap for buffer n
If we're still before MSC 1, then we'll ask to present at MSC
1. Otherwise, we'll ask to present at MSC 2.
Present already does this if the driver offers async flips, however it does
this by waiting for the right vblank event and sending an async flip right at
that point.
I've hacked the intel driver to offer this, but I get tearing at the top of
the screen. I think this is because flips are always done from within the
ring, and so the latency between the vblank event and the async flip happening
can cause tearing at the top of the screen.
That's why I'm keying the need for the extra buffer on the lack of 2D
driver support for async flips.
Signed-off-by: Keith Packard <keithp@keithp.com>
Acked-by: Jason Ekstrand <jason.ekstrand@intel.com>
Tested-by: Dylan Baker <baker.dylan.c@gmail.com>
Need to unwrap the indirect resource otherwise bad things will happen.
Fixes random crashes and timeouts with piglit's arb_indirect_draw tests.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
IVB had a restriction that prevented us from emitting compressed
three-source instructions, and although that was lifted on Haswell,
Haswell had a new restriction that said BFI instructions specifically
couldn't be compressed.
Transform
sqrt a, b
rcp c, a
into
sqrt a, b
rsq c, b
The improvement here is that we've broken a dependency between these
instructions. Leads to 330 fewer INV instructions and 330 more RSQ.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Transform
sqrt a, b
rcp c, a
into
sqrt a, b
rsq c, b
In most cases the sqrt's result is still used, so the improvement here
is that we've broken a dependency between these instructions. Leads to
80 fewer INV instructions and 80 more RSQ.
Occasionally the sqrt's result is no longer used, leading to:
instructions in affected programs: 5005 -> 4949 (-1.12%)
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
The next patch adds an algebraic optimization for the pattern
sqrt a, b
rcp c, a
and turns it into
sqrt a, b
rsq c, b
but many vertex shaders do
a = sqrt(b);
var1 /= a;
var2 /= a;
which generates
sqrt a, b
rcp c, a
rcp d, a
If we apply the algebraic optimization before CSE, we'll end up with
sqrt a, b
rsq c, b
rcp d, a
Applying CSE combines the RCP instructions, preventing this from
happening.
No shader-db changes.
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Helps a handful of programs in Serious Sam 3 that use do-while loops.
instructions in affected programs: 16114 -> 16075 (-0.24%)
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Do not rely on LLVMMCJITMemoryManagerRef being available.
The c binding to the memory manager objects only appeared
on llvm-3.4.
The change is based on an initial patch of Brian Paul.
Reviewed-by: Brian Paul <brianp@vmware.com>
Tested-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Mathias Froehlich <Mathias.Froehlich@web.de>
Blitter can still have transfers hanging around which it frees in
util_blitter_destroy(). So let it clean up before we yank the
transfer_pool from under it.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Indirect registers consume an additional token. Try to clean up the
token calculation math a bit, and fix it at the same time.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
If the name is just going to get dropped, don't bother making it. If
the name is made, release it sooner (rather than later).
No change Valgrind massif results for a trimmed apitrace of dota2.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
If the name is just going to get dropped, don't bother making it. If
the name is made, release it sooner (rather than later).
No change Valgrind massif results for a trimmed apitrace of dota2.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Valgrind massif results for a trimmed apitrace of dota2:
n time(i) total(B) useful-heap(B) extra-heap(B) stacks(B)
Before (32-bit): 74 40,578,719,715 67,762,208 62,263,404 5,498,804 0
After (32-bit): 52 40,565,579,466 66,359,800 61,187,818 5,171,982 0
Before (64-bit): 74 37,129,541,061 95,195,160 87,369,671 7,825,489 0
After (64-bit): 76 37,134,691,404 93,271,352 85,900,223 7,371,129 0
A real savings of 1.0MiB on 32-bit and 1.4MiB on 64-bit.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
These few places were using ir_var_auto for seemingly no reason. The
names were not added to the symbol table.
No change Valgrind massif results for a trimmed apitrace of dota2.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
No change Valgrind massif results for a trimmed apitrace of dota2.
v2: Minor rebase on _mesa_init_constants changes.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Later patches will give every ir_var_temporary the same name in release
builds. Adding a bunch of variables named "compiler_temp" to the symbol
table can only cause problems.
No change Valgrind massif results for a trimmed apitrace of dota2.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Specifically, ir_var_temporary variables constructed with a NULL name
will all have the name "compiler_temp" in static storage.
No change Valgrind massif results for a trimmed apitrace of dota2.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Valgrind massif results for a trimmed apitrace of dota2:
n time(i) total(B) useful-heap(B) extra-heap(B) stacks(B)
Before (32-bit): 44 40,577,049,140 68,118,608 62,441,063 5,677,545 0
After (32-bit): 71 40,583,408,411 67,761,528 62,263,519 5,498,009 0
Before (64-bit): 63 37,122,829,194 95,153,008 87,333,600 7,819,408 0
After (64-bit): 67 37,123,303,706 95,150,544 87,333,600 7,816,944 0
A real savings of 173KiB on 32-bit and no change on 64-bit.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
At least one of these pointers must be NULL, and we can determine which
will be NULL by looking at other fields. Use this information to store
both pointers in the same location.
If anyone can think of a better name for the union than "u", I'm all
ears.
Valgrind massif results for a trimmed apitrace of dota2:
n time(i) total(B) useful-heap(B) extra-heap(B) stacks(B)
Before (32-bit): 63 40,574,239,515 68,117,280 62,618,607 5,498,673 0
After (32-bit): 44 40,577,049,140 68,118,608 62,441,063 5,677,545 0
Before (64-bit): 53 37,126,451,468 95,150,256 87,711,304 7,438,952 0
After (64-bit): 63 37,122,829,194 95,153,008 87,333,600 7,819,408 0
A real savings of 173KiB on 32-bit and 368KiB on 64-bit.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Also move num_state_slots inside ir_variable_data for better packing.
The payoff for this will come in a few more patches.
No change Valgrind massif results for a trimmed apitrace of dota2.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
The payoff for this will come in a few more patches.
No change Valgrind massif results for a trimmed apitrace of dota2.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
warn_extension_index was moved to improve packing.
Valgrind massif results for a trimmed apitrace of dota2:
n time(i) total(B) useful-heap(B) extra-heap(B) stacks(B)
Before (32-bit): 73 40,580,476,304 68,488,400 62,796,151 5,692,249 0
After (32-bit): 73 40,575,751,558 68,116,528 62,618,607 5,497,921 0
Before (64-bit): 71 37,124,890,613 95,889,584 88,089,008 7,800,576 0
After (64-bit): 62 37,123,578,526 95,150,784 87,711,304 7,439,480 0
A real savings of 173KiB on 32-bit and 368KiB on 64-bit.
v2: Use the enum name with the bit-field and remove the extra casts.
Suggested by Ken.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com> [v1]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> [v1]
Also move the new warn_extension_index into ir_variable::data. This
enables slightly better packing.
Valgrind massif results for a trimmed apitrace of dota2:
n time(i) total(B) useful-heap(B) extra-heap(B) stacks(B)
Before (32-bit): 82 40,580,040,531 68,488,992 62,973,695 5,515,297 0
After (32-bit): 73 40,580,476,304 68,488,400 62,796,151 5,692,249 0
Before (64-bit): 65 37,124,013,542 95,892,768 88,466,712 7,426,056 0
After (64-bit): 71 37,124,890,613 95,889,584 88,089,008 7,800,576 0
A real savings of 173KiB on 32-bit and 368KiB on 64-bit.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
The payoff for this will come in the next patch.
No change Valgrind massif results for a trimmed apitrace of dota2.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
After compilation (and before linking) we can eliminate quite a few
built-in variables. Basically, any uniform or constant (e.g.,
gl_MaxVertexTextureImageUnits) that isn't used (with one exception) can
be eliminated. System values, vertex shader inputs (with one
exception), and fragment shader outputs that are not used and not
re-declared in the shader text can also be removed.
gl_ModelViewProjectMatrix and gl_Vertex are used by the built-in
function ftransform. There are some complications with eliminating
these variables (see the comment in the patch), so they are not
eliminated.
Valgrind massif results for a trimmed apitrace of dota2:
n time(i) total(B) useful-heap(B) extra-heap(B) stacks(B)
Before (32-bit): 46 40,661,487,174 75,116,800 68,854,065 6,262,735 0
After (32-bit): 50 40,564,927,443 69,185,408 63,683,871 5,501,537 0
Before (64-bit): 64 37,200,329,700 104,872,672 96,514,546 8,358,126 0
After (64-bit): 59 36,822,048,449 96,526,888 89,113,000 7,413,888 0
A real savings of 4.9MiB on 32-bit and 7.0MiB on 64-bit.
v2: Don't remove any built-in with Transpose in the name.
v3: Fix comment typo noticed by Anuj.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Suggested-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Acked-by: Anuj Phogat <anuj.phogat@gmail.com>
Cc: Eric Anholt <eric@anholt.net>
All built-in uniforms are supposed to be backed by some GL state. The
state_slots field describes this backing state.
This helped me track down a bug in a later patch.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Acked-by: Anuj Phogat <anuj.phogat@gmail.com>
Reuse the LLVMContext already allocated in llvmpipe_context
for draw_llvm if ppossible. This should decrease the memory
footprint of an llvmpipe context.
v2: Fix compile with llvm disabled.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Signed-off-by: Mathias Froehlich <Mathias.Froehlich@web.de>
This fixes the remaining problem with the recently introduced
global jit memory manager. This change again uses a memory manager
that is local to gallivm_state. This implementation still frees
the majority of the memory immediately after compilation.
Only the generated code is deferred until this code is no longer used.
This change and the previous one using private LLVMContext instances
I can now safely run several independent OpenGL contexts driven
by llvmpipe from different threads.
v3: Rebase on llvm-3.6 compile fixes.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Signed-off-by: Mathias Froehlich <Mathias.Froehlich@web.de>
This is one step to make llvmpipe thread safe as mandated by the OpenGL
standard. Using the global LLVMContext is obviously a problem for
that kind of use pattern. The patch introduces two LLVMContext
instances that are private to an OpenGL context and used for all
compiles. One is put into struct draw_llvm and the other
one into struct llvmpipe_context.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Signed-off-by: Mathias Froehlich <Mathias.Froehlich@web.de>
The big pile of patches I just pushed regresses about 25 piglit tests on
SNB. This fixes the regressions.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
The screen argument isn't actually used by lp_jit_screen_init() at this
time, but let's move the call so that we pass a valid pointer.
v2: don't leak screen if lp_jit_screen_init() fails.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
For cube resources, the array_size value should be 6. So handle
that case as we do for array texture resources. But assert that
array_size==6 just to be safe.
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
layers (aka array_size) should be 6 for cube textures so we don't need
to special-case it. But add an assertion just to be safe.
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Earlier in the function we assert layers==6 for PIPE_TEXTURE_CUBE so
there's no reason to special-case the pt.array_size = layers assignment.
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
The core sw primitive restart code is still around, because i965 uses it
in some cases, but there are no drivers that want it on all the time.
Reviewed-by: Rob Clark <robdclark@gmail.com>
The drivers not flagging primitive restart support are r300 swtcl, svga,
nv30, and vc4.
The point of primitive restart is to slightly reduce draw call overhead
for apps by batching multiple draws. If we do an extra pass to read the
index buffer and split back into multiple draws, we've entirely missed the
point. This is particularly bad for drivers that otherwise have hardware
IB reads, where the readback is probably uncached.
Reviewed-by: Rob Clark <robdclark@gmail.com>
On gen 7, the MRF was removed and we gained the ability to do send
instructions directly from the GRF. This commit enables that
functinoality for FB writes.
v2: Make handling of components more sane.
i965/fs: Force a high register for the final FB write
v2: Renamed the array for the range mappings and added a comment
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
If we are going to use LOAD_PAYLOAD operations to fill MRF registers, then
we will need this.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Previously, we were use the base_mrf parameter of fs_inst to store the MRF
location. In preparation for doing FB writes from the GRF, we now also
allow you to set inst->base_mrf to -1 and provide a source register.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Now that we have execution sizes, we can use that instead of the
dispatch width. This way it also works for 8-wide instructions in
SIMD16.
i965/fs: Make effective_width a variable instead of a function
i965/fs: Preserve effective width in constant propagation
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
This will, eventually, allow us to manage execution sizes of
instructions in a much more natural way from the fs_visitor level.
i965/fs: Explicitly set instruction execute size a couple of places
i965/blorp: Explicitly set instruction execute sizes
Since blorp is all 16-wide and nothing isn't, in general, very careful
about register width, we'll just set it all explicitly.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Now that we track both halves of a 16-wide vgrf, we no longer need to worry
about force_sechalf or force_uncompressed. The only real issue is if the
destination is too small.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
This commit fixes a bug in register coalesce that happens when one register
is moved to another the proper number of times but the channels are
re-arranged. When this happens, the previous code would happily coalesce
the registers regardless of the fact that the channel mappins were wrong.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Now that offset() can properly handle MRF registers, we can use an MRF
fs_reg and let offset() handle incrementing it correctly for different
dispatch widths. While this doesn't have any noticeable effect currently,
it does ensure that the destination register is 16-wide which will be
necessary later when we start detecting execution sizes based on source and
destination registers.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
This is actually the squash of a bunch of different changes. Individual
commit titles follow:
i965/fs: Always 2-align registers SIMD16 for gen <= 5
i965/fs: Use the register width when applying offsets
This reworks both byte_offset() and offset() to be more intelligent.
The byte_offset() function now supports offsets bigger than 32. The
offset() function uses the byte_offset() function together with the
register width and the type size to offset the register by the correct
amount.
i965/fs: Change regs_read to be in hardware registers
i965/fs: Change regs_written to be actual hardware registers
i965/fs: Properly handle register widths in LOAD_PAYLOAD
The LOAD_PAYLOAD instruction is a bit special because it collects a
bunch of registers (with possibly different widths) into a single
payload block. Once the payload is constructed, it's treated as a
single block of data and most of the information such as register widths
doesn't matter anymore. In particular, the offset of any particular
source register is the accumulation of the sizes of the previous source
registers.
i965/fs: Properly set writemasks in LOAD_PAYLOAD
i965/fs: Handle register widths in demote_pull_constants
i965/fs: Get rid of implicit register doubling in the allocator
i965/fs: Reserve enough registers for PLN instructions
i965/fs: Make sources and destinations interfere in 16-wide
i965/fs: Properly handle register widths in CSE
i965/fs: Properly handle register widths in register_coalesce
i965/fs: Properly handle widths in copy propagation
i965/fs: Properly handle register widths in VARYING_PULL_CONSTANT_LOAD
i965/fs: Properly handle register widths and odd register sizes in spilling
i965/fs: Don't waste a register on texture lookups for gen >= 7
Previously, we were waisting a register in SIMD16 mode because we could
only allocate registers in pairs. Now that we can allocate and address
odd-sized registers, let's get rid of this special-case.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Every register in i965 assembly implicitly has a concept of a "width".
Usually, this is derived from the execution size of the instruction.
However, when writing a compiler it turns out that it is frequently a
useful to have the width explicitly in the register and derive the
execution size of the instruction from the widths of the registers used in
it.
This commit adds a width field to fs_reg along with an effective_width()
helper function. The effective_width() function tells you how wide the
register effectively is when used in an instruction. For example, uniform
values have width 1 since the data is not actually repeated, but when used
in an instruction they take on the width of the instruction. However, for
some instructions (LOAD_PAYLOAD being the notable exception), the width is
not the same.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Just pass the visitor into is_copy_payload() and is_coalesce_candidate()
instead of a register size and the virtual_grf_sizes array. Among other
things, this makes the code more obvious because you don't have to figure
out where src_size came from.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Right now, this function is a no-op but it indicates that we intend to only
use the first half of the 16-wide register.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
This commit reworks copy propagation a bit to support propagating the
copying of partial registers. This comes up every time we have pull
constants because we do a pull constant read immediately followed by a move
to splat the one component of the out to 8 or 16-wide. This allows us to
eliminate the copy and simply use the one component of the register.
Shader DB results:
total instructions in shared programs: 5044937 -> 5044428 (-0.01%)
instructions in affected programs: 66112 -> 65603 (-0.77%)
GAINED: 0
LOST: 0
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
A switch statement is much easier to read/edit than a big giant or
statement.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
This splits emit_fb_writes into two functions: emit_fb_writes and
emit_single_fb_write. This reduces the amount of duplicated code in
emit_fb_writes and makes the register number fiddling less arcane.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Sometimes these show up in LOAD_PAYLOAD instructions and it's nice to be
able to see them.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Previously we disabled compact_virtual_grfs when dumping optimizations.
The idea here was to make it easier to diff the dumped shader because you
didn't have a sudden renaming. However, sometimes a bug is affected by
compact_virtual_grfs and, when this happens, you want to keep dumping
instructions with compact_virtual_grfs enabled. By turning it into an
optimization pass and dumping it along with the others, we retain the
ability to diff because you can just diff against the compact_virtual_grf
output.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
We also set the register width equal to the dispatch width. Right now,
this is effectively a no-op since we don't do anything with it. However,
it will be important once we add an actual width field to fs_reg.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Previously, if an instruction wrote to more than one register, we
implicitly assumed that it filled the entire register. We never hit this
before because the only time we did multi-register writes was things like
texturing which always wrote to all of the registers. However, with the
upcoming ability to do 16-wide instructions in SIMD8 and things of that
nature, we can have multi-register writes at offsets and we'll hit this.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Using a floating-point type doesn't usually cause hangs on my HSW, but the
simulator complains about it quite a bit.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
We have this wonderful offset() function for advancing registers, but we're
not using it. Using offset() allows us to do some sanity checking and
avoid manually touching fs_reg::reg_offset. In a few commits, we will make
offset do even more nifty things for us.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
The original vgrf splitting code was written with the assumption that vgrfs
came in two types: those that can be split into single registers and those
that can't be split at all It was very conservative and bailed as soon as
more than one element of a register was read or written. This won't work
once we start allowing a regular MOV or ADD operation to operate on
multiple registers. This rewrite allows for the case where a vgrf of size
5 may appropriately be split in to one register of size 1 and two registers
of size 2.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Acked-by: Matt Turner <mattst88@gmail.com>
Previously, we were generating the fast-clear shader from GLSL. The
problem is that fast clears require that we use a replicated write rather
than a regular write instruction. In order to get this we had a
complicated and somewhat fragile optimization pass that looked for places
where we can use a replicated write and used it. Since replicated writes
have a lot of restrictions, we only ever use them for fast-clear
operations.
This commit replaces the optimization pass with a function that just
generates the shader we want. This is a) less code, b) less fragile than
the optimization pass, and c) generates a more efficient shader.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Fixes GPUVM faults when running the piglit test "getteximage-formats
init-by-rendering" with R600_DEBUG=forcedma on SI.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
We are currently only dealing with depth-only or stencil-only resources
here, not with resources having both depth and stencil[0]. In both cases,
the tiling mode index is in the tile_mode field, not in the
stencil_tile_mode field.
[0] Add an assertion for that.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Add finalize_vertex_elements() to finalize ilo_ve_state. This fixes a
potential issue with URB entry allocation for VS and move the complexity of
gen6_3DSTATE_VERTEX_ELEMENTS() to the new function.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
The code was kind of mixed up what buffers were getting stored in the case
that a resolve bit was unset (which are set based on the GL state at draw
time) and the buffer wasn't actually bound. In particular, depth-only
rendering would store the color buffer contents, which happen to be
pointing at the depth buffer.
Thanks to clearing out the resolve bits for things we really can't
resolve, now I can drop the safety checks for buffer presence around the
actual stores.
Fixes 42 piglit tests.
My attempts to clarify the code with _compacted/_uncompacted prefixed
variables apparently failed. Hopefully this is clearer.
In any case, the previous code wasn't clear enough to gcc to let it
optimize division by a power of two into a shift. No problems now.
Also, the previous code (in the ADD case) didn't work on 32-bit x86, due
to complicated set of interactions best summed up as unsigned division
and compiler optimizations.
Tested-by: Mark Janes <mark.a.janes@intel.com>
We need to keep track if a state change other than frag/vert shader
state will trigger us to need a different shader variant, and if
necessary mark the appropriate shader state as dirty. Otherwise we will
forget to re-emit the shader state.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
This is for hw that needs to emulate some texture wrap modes (like
CLAMP) with some help from the shader.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Keep the existing function as a common helper. But this lets us move an
a2xx specific hack out of common code. And the PIPE_TEX_WRAP_CLAMP
emulation will require an a3xx specific hack. So rather than piling on
hacks, split this out.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
We just clamp the incoming texture coordinates. This breaks the lambda
calculation, but it gets the piglit tests to pass. This is the same
behavior as in i965.
One spot in the docs says that it's stored at a miplevel just beyond the
last miplevel, which was scary. But really, you just load it as the R
coordinate (which conflicts with cubemaps, but you don't do border
clamping on cubes).
We have to expose them for GL 2.0, but we just always return a value of 0.
We should be advertising 0 query bits instead of 64, but gallium doesn't
have plumbing for that yet. At least this stops the segfaults.
This may reduce register pressure and uniform counts. Drops a bunch of 0
uniform loads on vs-temp-array-mat4-index-col-row-wr.shader_test, which is
failing to register allocate.
It's not passing some of the piglit tests, because it looks like at small
miplevels some contents from surrounding faces are getting filtered in at
the corners. It does get 7 new tests passing.
In the other related fields, "0" refers to the size of the first miplevel,
while this is a field in a slice. The other implicit slices we have
(cubemap layers) don't vary in size compared to the first one.
Almost always, the MOV will get copy propagated out. Even if it doesn't,
it's probably better to just reload the uniform at next use (to reduce
register pressure) rather than try to save instruction count.
I was looking at this because in the presence of texturing (which calls
add_uniform() directly to get the uniform load forced into the
instruction) the c->uniform_contents indices don't match 1:1 with the
temporary qregs.
commit 4ed23fd broke creation of pbuffer surfaces, patch fixes
the failure, noticed when running chrome with '--use-gl=egl'.
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
When an instruction's result was consumed by multiple mov.sat
instructions, we would decide that we couldn't move the saturate
modifier because something else was using the result, even though it was
just another mov.sat!
total instructions in shared programs: 4275598 -> 4274842 (-0.02%)
instructions in affected programs: 75634 -> 74878 (-1.00%)
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
When we find a mov.sat, we search backwards. We might as well search
everything else backwards as well and potentially look at fewer
instructions.
This change enables the next patch.
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Get rid of the 'default' case (as suggestied by imirkin) so compiler
warns us about missing caps. Also add some caps that were missing until
now.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
4155d1c7 'st/mesa: drop dependence on API profile in st_init_extensions'
broke freedreno because somehow 'PIPE_CAP_MAX_VIEWPORTS' fell through
the cracks. Resulting that we reported zero viewports. So the state
tracker never bothered to give us any valid viewport!
Signed-off-by: Rob Clark <robclark@freedesktop.org>
This is the only guaranteed way get the patch level for llvm,
since the define cannot always be found in config.h depending
on the version of llvm or the build system used.
CC: 10.2 10.3 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Jonathan Gray <jsg@jsg.id.au>
Added back in 2009, with osmesa/GLU in mind. Unlikely to be working
any more since the removal of the static makefiles.
Cc: Brian Paul <brianp@vmware.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Rather than having double negatives -> disable-opencl, default=no
simply use enabled/disabled. It makes things a bit easier for the
reader and consistent throughout the file.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
The location of the egl driver(s) is matter that we should have
never exposed to the user. Currently the dri2 driver is built
into the libEGL loader, with the gallium based one soon to follow.
v2: Fold EGL_DRIVER_INSTALL_DIR within the makefiles. Suggested by Matt.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=80615
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
It is not used outside the render code. There are also too many details in it
that we do not want other components to access directly.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Remove emit_draw() and ILO_RENDER_DRAW indirections. With all emit functions
being direct now, ilo_render_estimate_size() and more can also be removed.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Add these new high-level functions
ilo_render_get_draw_dynamic_states_len()
ilo_render_emit_draw_dynamic_states()
ilo_render_get_rectlist_dynamic_states_len()
ilo_render_emit_rectlist_dynamic_states()
ilo_render_get_draw_surface_states_len()
ilo_render_emit_draw_surface_states()
for draw and rectlist state emission. They are implemented in the new
ilo_render_dynamic.c and ilo_render_surface.c.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
For all supported query types, we always emit a PIPE_CONTROL. Call
ilo_render_get_flush_len() for simplicity and clarity.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
ilo_render is based on ilo_builder. We should only care if the builder
buffers are invalidated, or if the hardware context is invalidated. Replace
ilo_render_invalidate() with flags by ilo_render_invalidate_builder() and
ilo_render_invalidate_hw().
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Both dynamic buffer and surface buffer are state buffers. We should not use
state buffer to refer to the former.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Move members of ilo_3d that still make sense to ilo_context. With ilo_3d
gone, rename functions whose names begin with ilo_3d to something more
appropriate.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
According to GLSL(4.2) and GLSL-ES (1.0, 3.0) spec, Structures must
have the same name to be considered same type. We currently ignore
the name check while checking if two records are same. This patch
fixes this.
Patch fixes failing tests in WebGL conformance test
'shaders-with-uniform-structs' when running Chrome on OpenGL ES.
v2: Do not force name comparison with unnamed types (Tapani)
v3: Cleanups (Matt)
Signed-off-by: Kalyan Kondapally <kalyan.kondapally@intel.com>
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83934
The draw module would still try to use gallivm, causing many piglit tests
to fail with an assertion failure. llvmpipe might have been similarly
affected.
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
mesa_texstore expects pixel data, not compressed data. For compressed
textures, we want to just copy the bits in without any conversion.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Acked-by: Francisco Jerez <currojerez@riseup.net>
What happens is that a SPLIT operation is part of the spill node, and as
a pseudo op, the instruction gets erased after processing its first def.
However the later defs still need to refer to it, so instead delay
deleting until after that whole RA node is done processing.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=79462
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.2 10.3" <mesa-stable@lists.freedesktop.org>
These are pretty catastrophic, "should never happen" failure paths (though
4 tests in piglit hit them currently, due to a single bug). An abort()
that you can gdb on easily is probably more useful than a clean exit,
particularly since a bug in piglit framework right now is causing early
exit(1)s to simply not be recorded in the results at all.
I only made IS_NEGATIVE(x) use signbit in commit 0f3ba405 in an attempt
to fix 54805, but it didn't help. We didn't use signbit on some
platforms and instead defined it to x < 0.0f.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Was only tracked to be printed at the end of configure, but configure
quits if it can't build something we requested, rather than silently
dropping it, so printing these directories has little use.
Note that I had to add support for testing the packed attribute to
m4/ax_gcc_func_attribute.m4.
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com> [C bits]
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Presumbly this will let clang and other compilers use the built-ins as
well.
Notice two changes specifically:
- in _mesa_next_pow_two_64(), always use __builtin_clzll and add a
static assertion that this is safe.
- in macros.h, remove the clang-specific definition since it should
be able to detect __builtin_unreachable in configure.
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com> [C bits]
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
The spec says the type must be W (JIP is 16-bits after all), but we've
been emitting it with a UD type all along and have experienced no
adverse effects. Changing the type to D allows ELSE and ENDIF
instructions to be compacted.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
We're currently emitting compactable control flow instruction the wrong
types, preventing their compaction. The next patch will fix this and
actually enable compaction.
On chips that cannot compact control flow instructions, attempts to find
a match in the datatype table will fail.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
The array was previously indexed in units of brw_compact_inst (8-bytes),
but before compaction all instructions are uncompacted, so every odd
element was unused.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Used to pass over previously compacted instructions in this loop, but no
longer. No point in checking.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
It may be possible to create a contrived example in which a 3-src
instruction would have been compacted on Gen < 8. I'd rather not
discover it in the wild.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Currently a no-op, since instruction compaction isn't implemented for the
generations that have a programmable strips-and-fans unit.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Despite what the Sandybridge PRM says, ENDIF has Jump Count in <dst>,
not JIP in <src1>. (The same mistake appears about WHILE as well).
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Both sizes are VERT_ATTRIB_MAX, so this has no effect. But it drops a
few trivial uses of the derived state.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Fredrik Höglund <fredrik@kde.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
I'm not familiar with this code, but this sure appears to be a typo.
It looks like the intent is to set each array element, not arrays[0]
each time. Notably, the loop just below uses "array", not "arrays".
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Fredrik Höglund <fredrik@kde.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Cc: mesa-stable@lists.freedesktop.org
The code in get.c that handles this uses ctx->Array.VAO->VertexAttrib,
which is a gl_vertex_attrib_array structure, not a gl_client_array.
The offsets of all fields happened to be the same in both structures, at
least on x86_64. "Size," "Type," and "Stride" are obviously the same:
both structures start with the same fields, in the same order.
"Enabled" is dicier: there are different fields before it in both
structures, including pointer sized values which might need special
alignment.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Fredrik Höglund <fredrik@kde.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Cc: mesa-stable@lists.freedesktop.org
Dead since the _MaxElement removal, but these functions seemed generally
applicable, so I decided to remove them in a separate patch.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Fredrik Höglund <fredrik@kde.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
max_index was coming from either the user telling us as part of
glDrawRangeElements, or from an incidental calculation as part of some
sort of primitive conversion fallback. Sometimes, it was just set to the
default "I don't know" ~0 value.
If it wasn't set to the actual max index, then the kernel would reject the
draw call for allowing out-of-bounds VBO reads. So, compute the max index
from the sizes of the VBOs, which isn't too expensive (unlike mapping and
reading the index buffer) and is reliable.
Fixes piglit vao-element-array-buffer.
Still some open questions.. and at any rate, no additional piglit passes
due to various wrap modes that we need to emulate in at least some
cases :-(
But it does fix some mystery page-faults.. So add some comments in the
code where there are things that we need to emulate or do more r/e, and
push as-is.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Instead of separate color and Z/S writemasks, just have one writemask
parameter that takes a mask of the PIPE_MASK_[RGBAZS] flags.
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
PIPE_TEX_MIPFILTER_x is not legal for the pipe_sampler_state::
min/mag_img_filter fields. But PIPE_TEX_MIPFILTER_x == PIPE_TEX_FILTER_x
so we were getting lucky.
This also makes the code consistent with u_blitter.c.
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
The register coalescing portion of this patch hurts three shaders in
Guacamelee by one instruction each, but examining the diff makes me
believe that what we were generating was (perhaps harmlessly) incorrect.
When the instructions aren't in a flat list, this wouldn't have worked.
Also, this should be faster.
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
The only trick is changing a break into a return true in register
coalescing, since the macro is actually a double loop, and break will do
something different than you expect. (Wish I'd realized that earlier!)
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Now that nothing invalidates the CFG, we can calculate_cfg() immediately
after emit_fb_writes()/emit_thread_end() and never again.
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
This cleans up the debug flags to be consistently indented, use bit
shifting instead of hex-values and fixes a bug where the new DEBUG_NO8 flag
used the same value as the DEBUG_VUE flag. This was hidden by the numbers not
being aligned. Also removes gaps in the range where DEBUG_IOCTL (0x4) and
DEBUG_REGION (0x400) used to be.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
LLVM commit r218316 removes the JITMemoryManager class, which is
the parent for a seemingly important class in gallivm. In order to
fix the build, I've wrapped most of lp_bld_misc.cpp in
if HAVE_LLVM < 0x0306 and modifyed the
lp_build_create_jit_compiler_for_module() function to return false
for 3.6 and newer which effectively disables the gallivm functionality.
I realize this is overkill, but I could not come up with a simple
solution to fix the build. Also, since 3.6 will be the first release
without the old JIT, it would be really great if we could
move gallivm to use the C API only for accessing MCJIT. There
is still time before the 3.6 release to extend the C API in
case it is missing some functionality that is required by gallivm.
This fixes heap corruption. The sampler view can be bound in the context,
so we cannot call destroy directly.
Reviewed-by: Brian Paul <brianp@vmware.com>
Instead, pass the layout of GS inputs in memory to the ES using the shader
key. Only 64 bits are needed to represent the layout in the key.
Mixing and matching different VS and GS shaders should now always work.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
I will need this for fixing sample shading with 1 sample.
The good news is that all shader pm4 states no longer use the current context
state, so we can generate the pm4 states outside of draw_vbo if needed.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Generic varyings in TGSI were based on the value of VARYING_SLOT_TEX0, so VAR0
was always GENERIC[22] (with tessellation patches). Some drivers might not
be able to cope with that.
This commit defines a proper mapping, so that PNTC is GENERIC[8] and VAR0 is
GENERIC[9].
Reviewed-by: Brian Paul <brianp@vmware.com>
The extensions and limits being set in the conditional block are core-only
anyway and don't have any effect on other profiles.
Reviewed-by: Brian Paul <brianp@vmware.com>
E.g. the 4.0 compatibility profile can be forced with:
MESA_GL_VERSION_OVERRIDE=4.0COMPAT
Some tests that I have require 4.0 compatibility.
Reviewed-by: Brian Paul <brianp@vmware.com>
Rather than duplicating the libdeps, extra define... all over the
targets, define them only once and use when applicable.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
The respective HAVE_{SOFT,LLVM}PIPE are already descriptive
enough. Additionally the svga modules does not really use either
one, but the auxiliary draw & gallivm modules.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
Rather than duplicating the libdeps, extra define... all over the
targets, define them only once and use when applicable.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
Rather than duplicating the libdeps, extra define... all over the
targets, define them only once and use when applicable.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Matt Turner <mattst88@gmail.com>
Rather than duplicating the libdeps, extra define... all over the
targets, define them only once and use when applicable.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
Rather than duplicating the libdeps, extra define... all over the
targets, define them only once and use when applicable.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
Rather than duplicating the libdeps, extra define... all over the
targets, define them only once and use when applicable.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
Rather than duplicating the libdeps, extra define... all over the
targets, define them only once and use when applicable.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
Rather than duplicating the libdeps, extra define... all over the
targets, define them only once and use when applicable.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
The dri, vdpau, omx, xvmc and gbm targets don't need any authentication
even the VL ones never used it. Either the respective loader or the
library itself (vl) is doing its auth prior to calling create_screen()
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
Make sure that MEGADRIVERS is set in order to create the hardlinks.
The variable name is not the most appropriate and will be sorted
out in upcoming commits.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
There are only 32 bits in the flatshade flags (which are 1 bit per
component), the simulator crashes when you use more than about this many
varyings, and the original Broadcom code drop only exposed 8 as well.
Fixes 26 piglit tests in the varying-packing group, and makes many others
go from crash to fail (due to not checking their varying counts and
treating link failures as failures). Regresses ARB_fp/minmax (due to 8
varyings instead of 10).
This is just the GL 1.1 flat shading of colors -- we don't need to support
TGSI constant interpolation bits, because we don't do GLSL 1.30.
Fixes 7 piglit tests.
They still provide register pressure since I haven't made a special class
for them, but since they're only live for one instruction it probably
doesn't matter.
This improves the readability of QPU assembly.
The A-file unpack is just like R4 unpack, except that if you don't do a
floating-point operation it won't do float conversion (so int16 gets
scaled up to int32).
This will let me more reliably allocate a-file registers, which are going
to be even more in demand when I start using a-file unpacks.
Also fixes a bug where the reservation of payload registers (FRAG_Z/W) was
off by one but just caused failure to register allocate at all if the
off-by-one was fixed.
The r300 gallium driver is using it outside of the Mesa tree, and I wanted
to do so for vc4 as well. Rather than make the multiple-definitions
problem even more complicated, just move it to more-shared code.
v2: Don't forget to delete the symlink in r300 (review by Matt).
Delete more r300-helper references (review by Emil)
Don't prefix util/ header inclusion with "util/" (review by Emil)
Reviewed-by: Matt Turner <mattst88@gmail.com> (v1)
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com> (v1)
ffeb77c7b0 had a typo which turned all signed
integer divisions into unsigned ones. Oops.
This gets us back the 51 little piglits
(all from glsl built-in-functions, fs/vs/gs-op-div-int-ivec2 and similar).
Cc: "10.2 10.3" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
If _mesa_get_tex_image() return NULL there is already error
set in context. Other error pats free allocated texture.
Signed-off-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Check if _mesa_meta_bind_rb_as_tex_image() did give the texture.
If no texture was given there is already either
GL_INVALID_VALUE or GL_OUT_OF_MEMORY error set in context.
Signed-off-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Running fast clear glClear with SNB caused Valgrind to
complain about this.
v2: line 237 fixed glClear from leaking memory, other
strdups are also now changed to ralloc_strdups but I
don't know what effect those have. At least no changes in
my Piglit quick run.
Signed-off-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Add current_pipe_control_dw1 and deferred_pipe_control_dw1 to track what have
been done since lsat 3DPRIMITIVE and what need to be done before next
3DPRIMITIVE. Based on them, we can emit WAs more smartly.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
It was used to set has_gen6_wa_pipe_control to false when the batch buffer
changed. When called from emit_flush() and others, it also unset
ILO_3D_PIPELINE_INVALIDATE_BATCH_BO so that the following emit_draw() will not
set has_gen6_wa_pipe_control to false again. It sounded error-prone and was
just ugly.
We should be able to achieve the same goal by reset has_gen6_wa_pipe_control
in ilo_3d_pipeline_invalidate(). With handle_invalid_batch_bo() gone, the
emit functions can also be inlined.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
We do not need to call it from GEN7 pipeline anymore since software
PIPE_QUERY_PRIMITIVES_EMITTED is gone.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Previously we would get a potentially computed post-swizzle coord based
on the texture target info, which would not include the bias/lod in the
last argument.
The second argument does not have to be adjacent, so adjusting the order
array did not make sense.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
This will make life a lot easier as we add support for additional
instructions.
v2: shadow reference value is always .z or .w
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
The only place the enum pipe_type was used is for the TGSI sampler
view return type. So make it a TGSI type. Note: it appears this
part of TGSI isn't used by anyone so it may be removed in the future.
v2: the new name is tgsi_return_type, not tgsi_type. This means we
can drop the previously posted tgsi_type -> tgsi_opcode_type patch.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Called when the user can insert new decls, instructions.
This could be used in a few places in the 'draw' module.
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
Instead we store a void pointer to the key, and cast it to
brw_wm_prog_key for fragment shader specific code paths.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
This helps:
1. Reduce the need to have fs_visitor::key's type be brw_wm_prog_key*
2. Align the code to allow brw_sampler_prog_key_data to be pulled out of other
prog_key types for different stages.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Instead we store a brw_stage_prog_data pointer, and cast it to
brw_wm_prog_data for fragment shader specific code paths.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Conditional rendering is not limited to draw_vbo(). Move the support to
ilo_context, and replace ilo_3d_pass_render_condition() by
ilo_skip_rendering().
SOL_RESET happens before bo execution. It should not be observed by the
commands that are already in the bo.
Move the code out of the pipeline now that it submits.
And config query and DRM_CONF_SHARE_FD to both mega-driver and
traditional build configs, so that EGL_EXT_image_dma_buf_import
works.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
We used different lists for different types of queries because we wanted to
update software queries quickly. Now that there is no software queries, we
are fine with a single list.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Read PIPE_QUERY_PRIMITIVES_GENERATED and PIPE_QUERY_PRIMITIVES_EMITTED from
hardware registers. Because all queries now have a bo, remove unnecessary
checks for q->bo.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Add support for PIPE_QUERY_PRIMITIVES_GENERATED and
PIPE_QUERY_PRIMITIVES_EMITTED in ilo_3d_pipeline_emit_query().
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
It replaces
ilo_3d_pipeline_emit_write_timestamp(),
ilo_3d_pipeline_emit_write_depth_count(), and
ilo_3d_pipeline_emit_write_statistics().
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Make it own()'s responsibility to make room for release() and itself. To be
able to do that, allow ilo_cp_submit() in own().
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Move pipe states in ilo_context to the new ilo_state_vector. The motivation
is that ilo_context consists of several loosely related things. When we need
an ilo_context somewhere, we usually need only one or two of the things in it.
This change makes ilo_state_vector one such thing.
An immediate result is that we no longer need ilo_context in 3D pipelines,
something we have planned for since early days.
Tested on my snb-gt2:
4 tests skip->pass in spec/EXT_texture_array
51 tests skip->pass in spec.glsl-3.30
4 tests skip->pass in spec/!OpenGL 3.3
No regressions; no skip->fail changes.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Some users don't understand that these variables can break OpenGL.
The general is rule is that if an app supports MSAA, you mustn't use
GALLIUM_MSAA.
For example, if an app has an 8xMSAA FBO and GALLIUM_MSAA=4
is set, resolving the FBO to the back buffer will be rejected which will look
like this on all gallium drivers:
http://www.phoronix.com/scan.php?page=article&item=amd_radeonsi_msaa
The environment variables also have no effect on modern apps like TF2, but
there is still a performance hit due to wasted bandwidth and VRAM.
In a nutshell, it does more harm than good.
Cc: 10.2 10.3 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
These instructions only have vex encodings, thus they can't be used without
avx. (Technically, one can still use avx-128 if avx isn't available because
the environment doesn't store the ymm registers, however I don't think llvm
can.)
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Geometry shaders was the only thing we needed to enable GLSL 1.50 and
OpenGL 3.2 in gen6.
v2: Layered clears do not work properly in gen6 with OpenGL 3.2. Kenneth
and Jordan realized that for this to work we also need
GL_AMD_vertex_shader_layered (which requires OpenGL 3.2, so it could not be
enabled before this patch), so we agreed to enable this together with
OpenGL 3.2 in this patch.
Signed-off-by: Samuel Iglesias Gonsalvez <siglesias@igalia.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
In gen6 we will use the geometry shader implementation from gen6_gs_visitor.cpp
and keep the implementation in brw_vec4_gs_visitor.cpp for gen7+. Notice that
gen6_gs_visitor inherits from brw_vec4_gs_visitor so it is not a completely
seprate implementation of geometry shaders.
Also, gen6 does not support multiple dispatch modes, its default operation mode
is equivalent to gen7's SINGLE mode, so select that in gen6 for consistency.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Uniforms declared as uniform blocks are stored in ubo surfaces and need to
be pulled from the geometry shader program so make sure we upload them first
and do the same for pull constants.
This fixes all piglit tests that use uniform blocks:
bin/shader_runner tests/spec/glsl-1.50/uniform_buffer/gs-*
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
For gen6 geometry shaders we use the first BRW_MAX_SOL_BINDINGS entries of the
binding table for transform feedback surfaces. However, vec4_visitor will
setup the binding table so that textures use the same space in the binding
table. This is done when calling assign_common_binding_table_offsets(0) as
part if its run() method.
To fix this clash we add a virtual method to the vec4_visitor hierarchy to
assign the binding table offsets, so that we can change this behavior
specifically for gen6 geometry shaders by mapping textures right after the
first BRW_MAX_SOL_BINDINGS entries.
Also, when there is no user-provided geometry shader, we only need to upload
the binding table if we have transform feedback, however, in the case of a
user-provided geometry shader, we can't only look into transform feedback
to make that decision.
This fixes multiple piglit tests for textureSize() and texelFetch() when these
functions are called from a geometry shader in gen6, like these:
bin/textureSize gs sampler2D -fbo -auto
bin/texelFetch gs usampler2D -fbo -auto
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Currently we buffer transform feedack varyings separately. This patch makes
it so that we reuse the values we have already buffered for all the output
varyings of the geometry shader instead.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Since geometry shaders can alter the value of varyings packed in the first
output VUE slot (PSIZ), we need to buffer it together with all the other
vertex data so we can emit the right value for each vertex when we do the
URB writes.
This fixes the following piglit test in gen6:
tests/spec/glsl-1.50/execution/redeclare-pervertex-out-subset-gs.shader_test
Signed-off-by: Samuel Iglesias Gonsalvez <siglesias@igalia.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Update gen6_gs_binding_table and gen6_sol_surface to use user-provided
geometry program information when present. This is necessary to implement
transform feedback support.
Signed-off-by: Samuel Iglesias Gonsalvez <siglesias@igalia.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
This takes care of generating code required to handle transform feedback.
Notice that transform feedback isn't enabled yet, since that requires
additional setups in other parts of the code that will come in later patches.
Signed-off-by: Samuel Iglesias Gonsalvez <siglesias@igalia.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
We will use this parameter in later patches to provide information relevant
to transform feedback that needs to be set as part of the FF_SYNC message.
Signed-off-by: Samuel Iglesias Gonsalvez <siglesias@igalia.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
This opcode generates code to copy the specified destination index
into subregister 5 of the MRF message header.
Signed-off-by: Samuel Iglesias Gonsalvez <siglesias@igalia.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
This opcode will be used when sending SVB WRITE messages to save
transform feedback outputs into Streamed Vertex Buffers.
Signed-off-by: Samuel Iglesias Gonsalvez <siglesias@igalia.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
So far in gen6 we only used geometry shaders to implement transform feedback
in vertex shaders, so we assumed that the VUE map for the geometry shader
stage was always the same as for the vertex shader stage. This is no longer
true now that we support user provided geometry shaders in gen6 too.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
For this we will need to move PrimitiveID information, delivered in the thread
payload in r0.1, to a separate register (we use GS_OPCODE_SET_PRIMITIVE_ID
for this), then map the corresponding varying slot to that register in the
setup_payload() method.
Notice that we cannot use a virtual register as the destination for the
PrimitiveID because we need to map all input attributes to hardware registers
in setup_payload(), which happens before virtual registers are mapped to
hardware registers. We could work around that issue if we were able to compute
the first non-payload register in emit_prolog() and move the PrimitiveID
information to that register, but we can't because at that point we still
don't know the final number uniforms that will be included in the payload.
So, what we do is to place PrimitiveID information in r1, which is always
delivered as part of the payload but its only populated with data
relevant for transform feedback when we set GEN6_GS_SVBI_PAYLOAD_ENABLE
in the 3DSTATE_GS state packet.
When we implement transform feedback, we wil make sure to move the value of r1
to another register before we overwrite it with the PrimitiveID.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
In gen6 the geometry shader payload includes the PrimitiveID information in
r0.1. When the shader code uses glPimitiveIdIn we will have to move this to
a separate hardware register where we can map this attribute. This opcode
takes the selected destination register and moves r0.1 there.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
In gen6 we need to end the thread differently depending on whether we have
emitted at least one vertex or not. In case we did, the EOT message must
always include the COMPLETE flag or else the GPU hangs. If we have not
produced any output, however, we can't use the COMPLETE flag.
This would lead us to end the program with an ENDIF opcode, which we want
to avoid (and actually is not permitted since it hits an assertion), so
instead what we do is that we always request a new VUE handle every time we do
an URB WRITE, even for the last vertex we emit. With this we make sure that
whether we have emitted at least one vertex or none at all we have to finish the
thread without writing to the URB, which works for both cases by setting the
COMPLETE and UNUSED flags in the EOT message.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Just in case the GS algorithm does not call EndPrimitive() for the last
primitive produced. This is relevant only for non point outputs, since for
this we are already setting the PrimEnd flag on each vertex we emit.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Geometry shaders in gen6 are significantly different from gen7+ so it is better
to have them implemented in a different file rather than adding gen6 branching
paths all over brw_vec4_gs_visitor.cpp.
This commit adds an initial implementation that only handles point output, which
is the simplest case.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
In gen7+ we emit vertices as they come, however in gen6 geometry shaders we
have to buffer vertex data for all vertices and then emit it all in one go
at the end. To achieve this we need to generalize emit_urb_slot() to store
vertex data in general purpose registers and not only MRF registers.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
We had GS_OPCODE_SET_DWORD_2_IMMED but this required its source argument to be
an immediate. In gen6 we need to set dword 2 of the URB write message header
from values stored in separate register, so we need something more flexible.
This change replaces GS_OPCODE_SET_DWORD_2_IMMED with GS_OPCODE_SET_DWORD_2.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Gen6 seems to require that EOT messages include the complete flag too or else
the GPU hangs. We add will this flag to the instruction when we emit the
thread end opcode.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Gen6 geometry shaders need to allocate URB handles for each new vertex they
emit after the first (the URB handle for the first vertex is obtained via the
FF_SYNC message).
This opcode adds the URB allocation mechanism to regular URB writes.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
This implements the FF_SYNC message required in gen6 geometry shaders to
get the initial URB handle.
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
This is needed to support user-provided geometry shaders, since the
brw_ff_gs_prog atom in gen6 only takes care of implementing transform feedback
for vertex shaders.
If there is no user-provided geometry shader the implementation falls back to
the original code.
Signed-off-by: Samuel Iglesias Gonsalvez <siglesias@igalia.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Currently, gen6 only uses geometry shaders for transform feedback so the state
we emit is not suitable to accomodate general purpose, user-provided geometry
shaders. This patch paves the way to add these support and the needed
3DSTATE_GS packet modifications for it.
Previous code that emitted state to implement transform feedback in gen6 goes
to upload_gs_state_adhoc_tf().
Signed-off-by: Samuel Iglesias Gonsalvez <siglesias@igalia.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Currently, when a geometry shader can't use dual object mode we fall back to
dual instance mode, however, when invocations == 1, single dispatch mode is
more performant and equally efficient in terms of register pressure.
Single dispatch mode requires that the driver can handle interleaving of
input registers, but this is already supported (dual instance mode has
the same requirement). However, to take full advantage of single dispatch mode
to reduce register pressure we would also need the ability to store two
separate vec4 output values into vec8 registers, which would approximately
double our capacity to store temporary values, but currently the vec4 visitor
and generator classes do not support this, so at the moment register pressure
in single and dual instance modes is the same.
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
It has been a bad name since we added the builder. Rename it to
ILO_DEBUG=batch to match i965, and call ilo_builder_decode() from
ilo_cp_submit_internal().
"Flush" is used for too many things already: pipe resource flush, pipe context
flush, pipe transfer region flush, and hardware pipeline flush. Rename it to
ilo_cp_submit(). As such, ILO_DEBUG=flush is renamed to ILO_DEBUG=submit.
Fredrik's implementation of ARB_vertex_attrib_binding introduced new
gl_vertex_attrib_array and gl_vertex_buffer_binding structures, and
converted Mesa's older gl_client_array to be derived state. Ultimately,
we'd like to drop gl_client_array and use those structures directly.
One hitch is that gl_client_array::_MaxElement doesn't correspond to
either structure (unlike every other field), so we'd have to figure out
where to store it. The _MaxElement computation uses values from both
structures, so it doesn't really belong in either place. We could put
it in the VAO, but we'd have to pass it around everywhere.
It turns out that it's only used when ctx->Const.CheckArrayBounds is
set, which is only set by the (rarely used) classic swrast driver.
It appears that drivers/x11 used to set it as well, which was intended
to avoid segmentation faults on out-of-bounds memory access in the X
server (probably for indirect GLX clients). However, ajax deleted that
code in 2010 (commit 1ccef926be).
The bounds checking apparently doesn't actually work, either. Non-VBO
attributes arbitrarily set _MaxElement to 2 * 1000 * 1000 * 1000.
vbo_save_draw and vbo_exec_draw remark /* ??? */ when setting it, and
the i965 code contains a comment noting that _MaxElement is often bogus.
Given that the code is complex, rarely used, and dubiously functional,
it doesn't seem worth maintaining going forward. This patch drops it.
This will probably mean the classic swrast driver may begin crashing on
out of bounds vertex buffer access in some cases, but I believe that is
allowed by OpenGL (and probably happened for non-VBO accesses anyway).
There do not appear to be any Piglit regressions, either.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Acked-by: Roland Scheidegger <sroland@vmware.com>
While depth test state is passed through the fragment shader as sideband,
data, the stencil test state has to be set by the fragment shader itself.
Many tests are still failing, but this gets most of hiz/ passing.
The SWZ instruction can have swizzle terms >4 (SWIZZLE_ZERO, SWIZZLE_ONE).
These swizzle terms caused a few assertions to fail.
This started happening after the commit "mesa: Actually use the Mesa IR
optimizer for ARB programs." when replaying some apitrace files.
A new piglit test (tests/asmparsertest/shaders/ARBfp1.0/swz-08.txt)
exercises this.
Cc: "10.3" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
This allows for introducing dead code eliminating of uniforms, copy
propagation of uniforms, and instruction rescheduling between instructions
that both read uniforms.
This is particularly important for outputs, where we try to MOV the whole
vec4 to the VPM, even if only 1-3 components had been set up. It might
also be important for temporaries, if the shader reads components before
writing them.
While the result of signed integer division by zero is undefined by glsl
(and doesn't exist with d3d10), we must not crash, so need to make sure we
don't get sigfpe much like udiv already does.
Unlike udiv where we return 0xffffffff (as required by d3d10) there is
no requirement right now to return anything specific so we use zero.
sample opcodes don't have valid texture target information (and I don't think
this should be changed), however it would be nice if we had that information
ready elsewhere, so stuff that information into the tgsi info when analyzing
a shader.
v2: Ilja Mirkin spotted some bugs wrt not handling msaa resources. So add them
and while there also add them to the tex opcode analysis this was cloned from
as well (plus get rid of some bug not detecting indirect textures there in some
cases too).
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
MESA_FORMAT_x8y8z8w8 puts the x channel in the least significant part of
the containing 32-bit integer, which is equivalent to PIPE_FORMAT_xyzw8888.
PIPE_FORMAT_x8y8z8w8 puts the x channel first in memory.
This patch fixes up the mesa<->gallium mapping accordingly.
Signed-off-by: Richard Sandiford <rsandifo@linux.vnet.ibm.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
MESA_FORMAT_LnAn puts the luminance in the least significant part of
the containing integer, which is equivalent to PIPE_FORMAT_LAnn.
PIPE_FORMAT_LnAn puts the luminance first in memory.
This patch fixes up the mesa<->gallium mapping accordingly.
Signed-off-by: Richard Sandiford <rsandifo@linux.vnet.ibm.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This means that each 8888 SRGB format has a reversed counterpart,
which is necessary for handling big-endian mesa<->gallium mappings.
v2: fix missing i965 additions. (Jason)
fix 127->255 max alpha for SRGB formats. (Jason)
v1: Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Signed-off-by: Richard Sandiford <rsandifo@linux.vnet.ibm.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The associated UNORM format already existed.
This means that each LnAn format has a reversed counterpart,
which is necessary for handling big-endian mesa<->gallium mappings.
[airlied: rebased onto current master]
Signed-off-by: Richard Sandiford <rsandifo@linux.vnet.ibm.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
...i.e. formats in which the first listed component is in the least
significant byte of the integer. The corresponding UNORM aliases already exist.
Signed-off-by: Richard Sandiford <rsandifo@linux.vnet.ibm.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This means that each RnGnBnxn format has a reversed counterpart,
which is necessary for handling big-endian mesa<->gallium mappings.
The associated UNORM and SRGB formats already exist.
Signed-off-by: Richard Sandiford <rsandifo@linux.vnet.ibm.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
...i.e. formats in which the alpha or green channel is first in memory.
This means that each LnAn and RnGn format has a reversed counterpart,
which is necessary for handling big-endian mesa<->gallium mappings.
Signed-off-by: Richard Sandiford <rsandifo@linux.vnet.ibm.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Jason pointed out the bug on review adding new formats,
but the existing format also appears to have the bug, so
use 255 as the max, these are SRGB no SNORM.
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Luminance is the least-significant byte of the uint16, rather than the
lowest byte in memory. Other parts of mesa already handle this correctly
for big-endian, and swrast already handles other MESA_FORMAT_x8y8 formats
correctly. This case was just an odd-one-out.
Signed-off-by: Richard Sandiford <rsandifo@linux.vnet.ibm.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Cc: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
MESA_FORMAT_R8G8B8X8_SNORM used a function called unpack_X8B8G8R8_SNORM
while MESA_FORMAT_R8G8B8X8_SRGB used a function called unpack_R8G8B8X8_SRGB.
This patch renames the SNORM function to have the same order as the
MESA_FORMAT name, like the SRGB function does.
Signed-off-by: Richard Sandiford <rsandifo@linux.vnet.ibm.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This was being shared using a ../../ get out of gallium into
mesa, and I swore when I did it I'd fix things when we got a util
dir, we did, so I have.
v2: move RGTC_DEBUG define
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This gets a ton of piglit working that crashes in waffle context
management stuff otherwise. Actually supporting mismatched FB sizes is at
best going to require some more load/store generals for color buffers, but
if I can't manage to do that I'll want to just have state_tracker reject
those FBOs as unsupported, rather than deny GL 2.1.
Previously, we would get a trailing ', ' which looked strange.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The goal here is to have an argument for the depth write opcode so that I
can do computed depth. In the process, this makes the calculations that
will be emitted more obvious in the QIR.
Commit afe3d1556f (i965: Stop doing
remapping of "special" regs.) stopped remapping delta_x/delta_y, and
additionally stopped considering them always-live. We later realized
delta_x was used in register allocaiton, so we actually needed to remap
it, which was fixed in commit 23d782067a
(i965/fs: Keep track of the register that hold delta_x/delta_y.).
However, that commit didn't restore the "always consider it live" part.
If all the code using delta_x was eliminated, fs_visitor::delta_x would
be left pointing at its old register number. Later code in register
allocation would handle that register number specially...even though it
wasn't actually delta_x.
To combat this, set delta_x/y to BAD_FILE if they're eliminated, and
check for that.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83127
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Cc: "10.3" <mesa-stable@lists.freedesktop.org>
The triangle_32_ rast functions never made it into the debug output,
confused me for a few seconds.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This patch builds on 6c8f547f66 and
previous patches by allowing u_format.csv to specify separate big-endian
and little-endian layouts. It then uses this to specify the correct layouts
for various depth/stencil formats. Later patches handle other formats.
To recap, the idea is that u_format.csv lists the channels for an N-byte
value as though it were an N-byte integer. For little-endian targets
the channels are listed starting at the least-significant bit of the
integer while for big-endian targets the channels are listed starting
at the most-significant bit. This means that for something like
PIPE_FORMAT_B8G8R8A8_UNORM (blue in first byte of memory, alpha in last
byte of memory) the orders are the same for both endiannesses. But for
something like PIPE_FORMAT_S8_UINT_Z24_UNORM, where the stencil is in
the least significant byte of a 32-bit integer, there need to be separate
channel definitions for each endianness.
The effect of this patch is to make the affected PIPE_FORMAT_*s have
the same layout as the associated MESA_FORMAT_*s for big-endian.
The MESA_FORMAT_*s are already handled correctly.
Fixes various piglit tests on z. No regressions on x86_64.
[airlied: squash subsequent patches]
util: Add big-endian layout for 5551 and 565 formats
util: Add big-endian layout for 10/10/10/2 formats
util: Add big-endian layout for 4444 formats
util: Add big-endian layout for 233 format
util: Add big-endian layout for 44 formats
Signed-off-by: Dave Airlie <airlied@redhat.com>
llvmpipe treats PIPE_FORMAT_Z32_FLOAT_S8X24_UINT as a bit of a special case,
handling it as two 32-bit pieces rather than a single 64-bit block:
/* 64bit d/s format is special already extracted 32 bits */
total_bits = format_desc->block.bits > 32 ? 32 : format_desc->block.bits;
The format_desc describes the whole 64-bit block, so the z shift
will be 32 for big-endian. But since we're accessing the z channel
as a 32-bit value rather than a 64-bit value, we need to mask the shift
with 31.
Signed-off-by: Richard Sandiford <rsandifo@linux.vnet.ibm.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This is just a very limited version, in particular sampler and sampler view
index must be the same. It cannot handle any modifiers neither.
Works much the same as soa version otherwise, to figure out the target we
need to store the sampler view dcls.
While here, also handle (no-op) RET and get rid of a couple bogus deprecated
comments.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
sample opcodes are a little oddly represented in the opcode_info, since
they don't count as texture instructions - they don't have valid target
information, but they may have offsets (unlike "ordinary" texture
instructions, the texture token may be optional for them).
So just make sure with these opcodes the optional offsets are accepted.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
sample opcodes don't encode a texture target, it would thus always
print UNKNOWN, which is not helpful (and wouldn't parse when giving
back the shader text to tgsi).
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
We don't have any specific limits in the hardware, just like the other
GPUs, so match their behavior. Fixes minmax_gles2 and several other
piglit tests relying on the specced uniform minmax values.
Put macro code in do {} while loop and put semicolons on macro calls
so auto indentation works properly.
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
This reduces gcc -O3 compile time to 1/4 of what it was on my system.
Reduces MSVC release build time too.
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
With GLES we don't give any kind of warning in case we don't
write to gl_position. This patch makes changes so that we
generate a warning in case of GLES (VER < 300) and an error
in case of GL.
Signed-off-by: Kalyan Kondapally <kalyan.kondapally@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Remap table for uniforms may contain empty entries when using explicit
uniform locations. If no active/inactive variable exists with given
location, remap table contains NULL.
v2: move remap table bounds check before existence check (Ian Romanick)
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Tested-by: Erik Faye-Lund <kusmabite@gmail.com> (v1)
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83574
We might run into ve->count == 0 and last_velement_edgeflag == true in
gen6_3DSTATE_VERTEX_ELEMENTS() when the state tracker sets an invalid
combination of VS and VE (does not seem to happen with st/mesa). Do not
assume ve->count is positive when last_velement_edgeflag is true.
Reported by Coverity.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Rename ilo_builder_batch_state_base_address() to gen6_state_base_address() for
consistency and remove unused gen6_STATE_BASE_ADDRESS(). Reorder the code in
gen6_PIPE_CONTROL() a bit. Finally, some mostly cosmetic changes.
Move functions for the 3D pipeline to the new headers. We artificially split
the functions into top (vertex processing) and bottom (pixel processing), to
keep the headers at reasonable sizes.
ir_rvalue::constant_expression_value() recursively walks down an IR
tree, attempting to reduce it to a single constant value. This is
useful when you want to know whether a variable has a constant
expression value at all, and if so, what it is.
The constant folding optimization pass attempts to replace rvalues with
their constant expression value from the bottom up. That way, we can
optimize subexpressions, and ideally stop as soon as we find a
non-constant subexpression.
In order to obtain the actual value of an expression, the optimization
pass calls constant_expression_value(). But it should only do so if it
knows the value can be combined into a constant. Otherwise, at each
step of walking back up the tree, it will walk down the tree again, only
to discover what it already knew: it isn't constant.
We properly avoided this call for ir_expression nodes, but not for
ir_swizzle nodes. This patch fixes that, drastically reducing compile
times on certain shaders where tree grafting has given us huge
expression trees. It also fixes SuperTuxKart.
Thanks to Iago and Mike for help in tracking this down.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=78468
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Cc: mesa-stable@lists.freedesktop.org
The FS backend has always used 0, and the VS backend has always used 1.
I think 1 is just working around other problems, and is incorrect.
Samplers are baked in; nothing uses the UNIFORM register we would
create, and we shouldn't upload any constant values for them.
Fixes ES3-CTS.shaders.struct.uniform.sampler_array_vertex.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Tested-by: Ian Romanick <ian.d.romanick@intel.com>
Samplers take up zero slots and therefore don't exist in the params
array, nor are they included in stage_prog_data->nr_params. There's no
need to store their size in param_size, as it's only used for dealing
with arrays of "real" uniforms (ones uploaded as shader constants).
We run into all kinds of problems trying to refer to the uniform storage
for variables that don't have uniform storage. For one, we may use some
other variable's index, or access out of bounds in arrays. In the FS
backend, our extra 2 * MaxSamplerImageUnits params for texture rectangle
rescaling paper over a lot of problems. In the VS backend, we claim
samplers take up a slot, which also papers over problems.
Instead, just skip allocating storage for variables that don't have any.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Tested-by: Ian Romanick <ian.d.romanick@intel.com>
We always uploaded them together, mostly out of laziness - both required
an additional vertex element. However, gl_VertexID now also requires an
additional vertex buffer for storing gl_BaseVertex; for non-indirect
draws this also means uploading (a small amount of) data. This is extra
overhead we don't need if the shader only uses gl_InstanceID.
In particular, our clear shaders currently use gl_InstanceID for doing
layered clears, but don't need gl_VertexID.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "10.3" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Tested-by: Ian Romanick <ian.d.romanick@intel.com>
In the non-indirect draw case, we call intel_upload_data to upload
gl_BaseVertex. It makes brw->draw.draw_params_bo point to the upload
buffer, and increments the upload BO reference count.
So, we need to unreference it when making brw->draw.draw_params_bo point
at something else, or else we'll retain a reference to stale upload
buffers and hold on to them forever.
This also means that the indirect case should increment the reference
count on the indirect draw buffer when making brw->draw.draw_params_bo
point at it. That way, both paths increment the reference count, so
we can safely unreference it every time.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "10.3" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Tested-by: Ian Romanick <ian.d.romanick@intel.com>
4f338c9b introduced logic to trigger a flush rather than overflowing
cmdstream buffer. But the threshold was too low, triggering flushes
where they were not needed. This caused problems with games like
xonotic.
Part of the problem is that we need to mark all state dirty between
cmdstream submit ioctls, because we cannot rely on state being
preserved across ioctls. But even with that, there are still some
problems that are still being debugged. For now:
1) correctly mark all state dirty
2) introduce FD_MESA_DEBUG flush flag to force rendering to be flushed
between each draw, to trigger problems (so that I can debug)
3) use a more reasonable threshold so for normal usecases we don't
trigger the problems
This at least corrects the regression, but there is still more debugging
to do.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
We need the .w component to end up in .x, since the hw appears to fetch
gl_FragColor starting with the .x coordinate regardless of MRT format.
As long as we are doing this, we might as well throw out the remaining
unneeded components.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Because of render-to-alpha (000x) shenanigans, freedreno needs to do
some special handling when rendering to alpha-only formats. And I
noticed that while we had _is_luminance(), _is_intensity(), etc, an
_is_alpha() helper was missing. So fix that.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Unreference the ctx->_Shader object before we delete all the pipeline
objects in the hash table. Before, ctx->_Shader could point to freed
memory when _mesa_reference_pipeline_object(ctx, &ctx->_Shader, NULL)
was called.
Fixes crash when exiting the piglit rendezvous_by_location test on
Windows.
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
q_total should never go below 0 (which is why it's defined as unsigned),
and if it does, then something is seriously wrong.
Signed-off-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
As noted in the previous commit, this was introduced in
567e2769b8 ("ra: make the p, q test more
efficient"), but I forgot to mention it.
Signed-off-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
In commit 567e2769b8 ("ra: make the p, q
test more efficient") I unknowingly introduced a new requirement to the
register allocator API: the user must set the register class of all
nodes before setting up their interferences, because
ra_add_conflict_list() now uses the classes of the two interfering
nodes. i965 already did this, but r300g was setting up register classes
interleaved with setting up the interference graph. This led to us
calculating the wrong q total, and in certain cases
e78a01d5e6 (" ra: optimistically color
only one node at a time") made it so that this bug caused a segfault. In
particular, the error occurred if the q total was decremented to 1 below
0 for the last node to be pushed onto the stack. Since q_total is an
unsigned integer, it overflowed to 0xffffffff, which is what
lowest_q_total happens to be initialzed to. This means that we would
fail the "new_q_total < lowest_q_total" check on line 476 of
register_allocate.c, and so the node would never be pushed onto the
stack, which led to segfaults in ra_select() when we failed to ever give
it a register.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=82828
Cc: "10.3" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Connor Abbott <cwabbott0@gmail.com>
Tested-by: Pavel Ondračka <pavel.ondracka@email.cz>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Needed for assert.
Fixes build on BE archs with -Werror=implicit-function-declaration.
In file included from
../../../../../src/gallium/auxiliary/draw/draw_fs.c:30:0:
../../../../../src/gallium/auxiliary/util/u_math.h: In function
'util_memcpy_cpu_to_le32':
../../../../../src/gallium/auxiliary/util/u_math.h:810:4: error:
implicit declaration of function 'assert'
[-Werror=implicit-function-declaration]
assert(n % 4 == 0);
^
Cc: "10.3" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Andreas Boll <andreas.boll.dev@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
In buf_clear_region() and buf_copy_region(), max_cmd_size was set to 0. If
either of the functions is called and there is not enough space in the
builder, the next ilo_cp_flush() will fail silently in a release build.
Replace magic numbers by size defines in tex_clear_region()/tex_copy_region()
for consistency and readability.
Intruduce gen6_blt_bo and gen6_blt_xy_bo to describe BOs. In the extreme case
of gen6_XY_SRC_COPY_BLT(), the number of parameters goes down from 18 to 8.
This allows a sampler view to have a different texture target than the
underlying resource. This will be used to implement the type casting
between 2d arrays and cube maps as specified in ARB_texture_view.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Fixes a regression from "nouveau/vdec: small fixes to h264 handling"
New picking order for frames:
1. Vidbuf pointer matches.
2. Take the first kicked ref.
3. If that fails, take a ref that has a different last_used.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Cc: "10.2 10.3" <mesa-stable@lists.freedesktop.org>
The BSP bo might be too small to contain all of the bsp data,
bump its size on overflow. Also bump inter_bo when this happens,
it might be too small otherwise.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Cc: "10.2 10.3" <mesa-stable@lists.freedesktop.org>
If the source is not a GRF, it could have a register >= virtual_grf_count.
Accessing virtual_grf_end with such a register would lead to
out-of-bounds access. Make sure the source is a GRF before accessing
virtual_grf_end.
Fixes Valgrind complaints while compiling some shaders.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: mesa-stable@lists.freedesktop.org
If the glx/wgl state tracker requested a core profile but the gallium
driver did not support some feature of GL 3.1 or later, we were setting
ctx->Version=0 and then failing the assertion in
_mesa_initialize_exec_table().
With this change we check for ctx->Version=0 and tear down the context
and return NULL from st_create_context().
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
So far we have been using CL_INVOCATION_COUNT to resolve this query but this
is no good with streams, as only stream 0 reaches the clipping stage.
From ARB_transform_feedback3:
"When a generated primitive query for a vertex stream is active, the
primitives-generated count is incremented every time a primitive emitted to
that stream reaches the Discarding Rasterization stage (see Section 3.x)
right before rasterization. This counter is incremented whether or not
transform feedback is active."
Unfortunately, we don't have any registers that provide the number of primitives
written to a specific stream other than the ones that track the number of
primitives written to transform feedback in the SOL stage, so we can't
implement this exactly as specified.
In the past we implemented this feature by activating the SOL unit even if
transform feeback was disabled, but making it so that all buffers were
disabled and it only recorded statistics, which gave us the right semantics
(see 3178d2474a). Unfortunately, this came with
a significant performance impact and had to be reverted.
This new take does not intend to implement the exact semantics required by
the spec, but improves what we have now, since now we return the primitive
count for stream 0 in all cases. With this patch we use
GEN7_SO_PRIM_STORAGE_NEEDED to resolve GL_PRIMITIVES_GENERATED queries
for non-zero streams. This would return the number of primitives written
to transform feedback for each stream instead. Since non-zero streams are
only useful in combination with transform feedback this should not be too
bad, and the only case that I think we would not be supporting would be
the one in which we want to use both GL_PRIMITIVES_GENERATED and
GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN on the same non-zero stream to
detect buffer overflow.
This patch also fixes the following piglit test:
arb_gpu_shader5-xfb-streams-without-invocations
This test uses both GL_PRIMITIVES_GENERATED and
GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN queries on non-zero streams, but it
does never hit the overflow case, so both queries are always expected to return
the same value.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "10.3" <mesa-stable@lists.freedesktop.org>
That better matches the actual userspace use case, the
kernel will force it to VRAM if the hardware requires it.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Less CPU overhead and avoids contention over CPU accessible memory on startup.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
In preparation to using buffers clears with the hw engine(s).
v2: split out flipping to using hw buffer clears.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
This builds the opengl DLLs with address layout space randomization
(ASLR) and data execution prevention (DEP) for better security.
Reviewed-by: Kurt Daverman <krd@vmware.com>
The old disassembler was modified from i965's. It is as much work as doing a
new one to keep it up-to-date, which also requires copying more headers over.
The outputs of this new disassembler should match i965's as closely as
possible.
Add some new registers and some tweaks. The changes that affect ilo are
GEN6_REG_HS_INVOCATION_COUNT -> GEN7_REG_HS_INVOCATION_COUNT
GEN6_REG_DS_INVOCATION_COUNT -> GEN7_REG_DS_INVOCATION_COUNT
GEN6_COND_NORMAL -> GEN6_COND_NONE
If a precision qualifer is allowed on type T, it should be allowed
on an array of T. Refactor the check to ensure this is the case.
(Fixes failures in WebGL conformance test 'gl-min-textures')
Signed-off-by: Frank Henigman <fjhenigman@google.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
MSVC replaces the "F" in "255.0F" with the macro argument which leads
to an error. s/F/FLT/ to avoid that.
It turns out we weren't using this macro at all on MSVC until the
recent "mesa: Drop USE_IEEE define." change.
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
This patch fixes a build error on DragonFly.
CC libpipe_loader_la-pipe_loader_drm.lo
pipe_loader_drm.c: In function 'pipe_loader_drm_probe':
pipe_loader_drm.c:207:10: error: implicit declaration of function 'close' [-Werror=implicit-function-declaration]
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Apparently guardband clipping doesn't work like we thought: objects
entirely outside fthe guardband are trivially rejected, regardless of
their relation to the viewport. Normally, the guardband is larger than
the viewport, so this is not a problem. However, when the viewport is
larger than the guardband, this means that we would discard primitives
which were wholly outside of the guardband, but still visible.
We always program the guardband to 8K x 8K to enforce the restriction
that the screenspace bounding box of a single triangle must be no more
than 8K x 8K. So, if the viewport is larger than that, we need to
disable guardband clipping.
Fixes ES3 conformance tests:
- framebuffer_blit_functionality_negative_height_blit
- framebuffer_blit_functionality_negative_width_blit
- framebuffer_blit_functionality_negative_dimensions_blit
- framebuffer_blit_functionality_magnifying_blit
- framebuffer_blit_functionality_multisampled_to_singlesampled_blit
v2: Mention the acronym expansion for TA/TR/MC in the comments.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Now that we have the data available, we need to expose it to the
shaders. We can reuse the same vertex element that we use for
gl_VertexID, but we need to back it by an actual vertex buffer.
A hardware restriction requires that vertex attributes coming from a
buffer (STORE_SRC) must come before any other types (i.e. STORE_0).
So, we have to make gl_BaseVertex be the .x component of the vertex
attribute. This means moving gl_VertexID to a different component.
I chose to move gl_VertexID and gl_InstanceID to the .z and .w
components, respectively, to make room for gl_BaseInstance in the .y
component (which would also come from a buffer, and therefore be
STORE_SRC).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
We'll need to emit another VERTEX_BUFFER_STATE for gl_BaseVertex;
pulling this into a helper function will save us from having to deal
with cross-generation differences in that code.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
This will be used for GL_ARB_shader_draw_parameters, as well as fixing
gl_VertexID, which is supposed to include gl_BaseVertex's value.
For indirect draws, we simply point at the indirect buffer; for normal
draws, we upload the value via the upload buffer.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
The lower_vertex_id pass converts uses of the gl_VertexID system value
to the gl_BaseVertex and gl_VertexIDMESA system values. Since
gl_VertexID is no longer accessed, it would not be considered active.
Of course, it should be, since the shader uses gl_VertexID.
v2: Move the var->name dereference past the var != NULL check.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Converts gl_VertexID to (gl_VertexIDMESA + gl_BaseVertex). gl_VertexIDMESA
is backed by SYSTEM_VALUE_VERTEX_ID_ZERO_BASE, and gl_BaseVertex is backed
by SYSTEM_VALUE_BASE_VERTEX.
v2: Put the enum in struct gl_constants and propoerly resolve the scope
in C++ code. Fix suggested by Marek.
v3: Reabase on Matt's foreach_in_list changes (was using foreach_list).
v4 (Ken): Use a systemvalue instead of a uniform because
STATE_BASE_VERTEX has been removed.
v5: Use a boolean to select lowering, and only allow one lowering
method. Suggested by Ken.
v6 (Ken): Replace strcmp against literal "gl_BaseVertex"/"gl_VertexID"
with SYSTEM_VALUE enum checks, for efficiency.
v7: Rebase on context constant initialization work.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
There exists hardware, such as i965, that does not implement the OpenGL
semantic for gl_VertexID. Instead, that hardware does not include the
value of basevertex in the gl_VertexID value.
SYSTEM_VALUE_VERTEX_ID_ZERO_BASE is the system value that represents
this semantic.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
v2: Additions to the documentation for SYSTEM_VALUE_VERTEX_ID. Quote
the GL_ARB_shader_draw_parameters spec and mention DirectX SV_VertexID.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
181581280b changed the way the
llvm-config version is read from sed to grep and introduced
a requirement for gnu grep extension that treats BREs as EREs.
Avoid this by calling egrep instead of grep which should be
able to handle EREs everywhere.
This allows Mesa to build on OpenBSD again.
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
This doesn't quite make depth-tex-compare work, presumably because we're
not hitting equality with itof(sample) * 1.0/0xffffff in the 0xffffff
case. arb_fragment_program_shadow tests pass, though, as well as a bunch
of other shadow-related stuff.
We potentially need to be careful that use of a value stored in r4 isn't
copy-propagated (or something) across another r4 write. That doesn't
appear to happen currently, and this makes the dataflow more obvious. It
also opens up not unpacking the r4 value, which will be useful for depth
textures.
Since software primitive-restart emulation is going to be removed (and
anyways, mostly seemed to be crash prone in combination with
u_primconvert and oddball scenarios (like PIPE_PRIM_POLYGON with only a
single vertex), might as well do it in hardware (which fortunately
didn't turn out to be too hard to figure out).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Triggered by shaders like:
FRAG
PROPERTY FS_COLOR0_WRITES_ALL_CBUFS 1
DCL OUT[0], COLOR
DCL CONST[0]
DCL TEMP[0..2], LOCAL
0: IF CONST[0].xxxx :0
1: MOV TEMP[0], TEMP[1]
2: ELSE :0
3: MOV TEMP[0], TEMP[2]
4: ENDIF
5: MOV OUT[0], TEMP[0]
6: END
not really a sane shader, although driver segfaulting is probably
not the appropriate response.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
We currently aren't too clever about dealing with running out of
cmdstream buffer space. Since we use a single buffer for both drawing
and tiling commands, we need to ensure there is enough space at the tail
of the cmdstream buffer to fit the tiling commands.
Until we get more clever, the easy solution is a threshold to trigger
flushing rendering even if the application does not trigger flush (swap,
changing render target, etc). This way we at least don't crash for apps
that do several thousand draw calls (like some piglit tests do).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Most of the things the new compiler still has trouble with basically
amount to cp stage removing too many copies. But without the cp stage,
the shaders the new compiler produces are still better (perf and
correctness) than the old compiler. So a simple thing to do until I
have more time to work on it is first trying falling back to new
compiler without cp, before finally falling back to old compiler.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Excess conversions considered harmful.
Recently Matt reworked the boolean uniform handling to use the value of
UniformBooleanTrue, rather than integer 1, when uploading uniforms:
mesa: Upload boolean uniforms using UniformBooleanTrue.
glsl: Use UniformBooleanTrue value for uniform initializers.
Marek then set the default to 1.0f for drivers without native integer
support:
mesa: set UniformBooleanTrue = 1.0f by default
However, ir_to_mesa was assuming a value of integer 1, and arranging for
it to be converted to 1.0f on upload. Since Marek's commit, we were
uploading 1.0f = 0x3f800000 which was being interpreted as the integer
value 1065353216 and converted to float as 1.06535322E9, which broke
assumptions in ir_to_mesa that "true" was exactly 1.0f.
+13 Piglits on classic swrast (fs-bool-less-compare-true,
{vs,fs}-op-not-bool-using-if, glsl-1.20/execution/uniform-initializer).
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83573
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Mesa already defines _GNU_SOURCE for glibc based systems and defining
_GNU_SOURCE will break the Mesa build on other systems such as OpenBSD.
_GNU_SOURCE only seems to be included in llvm-config output when
LLVM is built via autoconf and not when it is built by cmake.
Cc: "10.2 10.3" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Superseded by HAVE_LOADER_GALLIUM. The latter has a *DRM* brethren
making the whose easier on which one to keep.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Just drop the conditional and simplify our build. This means that
it'll build every time, but it does not require any dependencies nor
does it take that long to compile 200 lines of boilerplate code.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
The variable was unused and gave false information. The need for nonnull
winsys currently does not relate as it used to. Nowadays one can mix and
match more freely with plenty of winsys' to make your head spin.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
With recent commit we removed the NEED_NONNULL_WINSYS checks when
selecting the hardware (inc svga) winsys. svga has only one winsys
that explicitly requires libdrm (via it's bundled version of
vmwgfx_drm.h) but configure.ac never really checks for it.
Add the check early to prevent people from shooting themselves when
they select the driver but lack libdrm.
$ ./autogen.sh --disable-dri --disable-egl --disable-gallium-llvm
--with-dri-drivers=swrast --with-gallium-drivers=svga,swrast
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=82539
Cc: "10.2 10.3" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
The rest of stencil handling isn't done yet, but it documents an extra
cl_u8(0) and helps make it obvious why we don't need to format clear_depth
the same way the depth/stencil buffer is formatted.
For now it still requires the color buffer to be present -- we're relying
on the store of color buffer contents to end the frame, and we have to do
something with color buffers in the rendering config packet.
According to GLSL-ES Spec(i.e. 1.0, 3.0), gl_Position value is undefined
after the vertex processing stage if we don't write gl_Position. However,
GLSL 1.10 Spec mentions that writing to gl_Position is mandatory. In case
of GLSL-ES, it's not an error and atleast the linking should pass.
Currently, Mesa throws an linker error in case we dont write to gl_position
and Version is less then 140(GLSL) and 300(GLSL-ES). This patch changes
it so that we don't report an error in case of GLSL-ES.
Signed-off-by: Kalyan Kondapally <kalyan.kondapally@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83380
Similar to the changes to GEN7 command functions, but to GEN6 this time.
As every GPE function has been converted, remove
ilo_cp_assert_no_implicit_flush() calls.
Make these changes
ilo_cp_begin() -> ilo_builder_batch_pointer()
ilo_cp_write() -> direct memory set
ilo_cp_write_bo() -> ilo_builder_batch_reloc()
and use this chance to drop the "_emit_" infix.
Make these changes
ilo_cp_steal_ptr() and memcpy() -> ilo_builder_state_write()
ilo_cp_steal_ptr() -> ilo_builder_state_pointer()
and use this chance to drop the "_emit_" infix.
Make these changes
ilo_cp_steal_ptr() and memcpy() -> ilo_builder_surface_write()
ilo_cp_steal() and ilo_cp_write() -> ilo_builder_surface_write()
ilo_cp_write_bo() -> ilo_builder_surface_reloc()
and use this chance to drop the "_emit_" infix.
Make these changes
ilo_cp_begin() -> ilo_builder_batch_pointer()
ilo_cp_write() -> direct memory set
ilo_cp_write_bo() -> ilo_builder_batch_reloc()
and make sure there is no implicit flush. Use this chance to drop the
"_emit_" infix.
Remove instruction buffer management from ilo_3d and adapt ilo_shader_cache to
upload kernels to ilo_builder. To be able to do that, we also let ilo_builder
manage STATE_BASE_ADDRESS.
Comparing to how we manage batch and instruction buffers, the new builder
- does not flush
- manages both types of buffers
- manages STATE_BASE_ADDRESS
- uploads kernels using unsynchronized mapping
- has its own decoder for the buffers
- provides more helpers
Do not require a toy_compiler so that it can be used in other places, such as
state dumping. Add a bool to control whether the raw instruction words are
shown.
This would just crash. Noticed by accident while checking int divisions by zero
with a quickly hacked piglit test.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Forgot to add it when I fixed up the start instance handling in (llvm) draw.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Calling the variable min when it's really max and vice versa seems a bit
confusing.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
UBO loads can be boolean-valued expressions, too, so we need to handle
them in emit_bool_to_cond_code() and emit_if_gen6().
However, unlike most expressions, it doesn't make sense to evaluate
their operands, then do something with the results. We just want to
evaluate the UBO load as a whole---which performs the read from
memory---then load the boolean result into the flag register.
Instead of adding code to handle it, we can simply bypass the
ir_expression handling, and fall through to the default code, which will
do exactly that.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83468
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Cc: mesa-stable@lists.freedesktop.org
Matt and I believe that Sandybridge actually uses 0xFFFFFFFF for a
"true" comparison result, similar to Ivybridge. This matches the
internal documentation, and empirical results, but contradicts the PRM.
So, the comment is inaccurate, and we can actually just handle these
directly without ever needing to fall through to the condition code
path.
Also, the vec4 backend has always done it this way, and has apparently
been working fine. This patch makes the FS backend match the vec4
backend's behavior.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
ir_triop_csel can return a boolean expression, so we need to handle it
here; we simply forgot when we added ir_triop_csel, and forgot again
when adding it to emit_bool_to_cond_code.
Fixes Piglit's EXT_shader_integer_mix/{vs,fs}-mix-if-bool on Sandybridge.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Cc: mesa-stable@lists.freedesktop.org
As long as we don't have a workaround for frame based
decoding in VDPAU we should not advertise NV_vdpau_interop.
v2: fix commit message, check if get_video_param is present
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: mesa-stable@lists.freedesktop.org
Instead we cast backend_visitor::prog for fragment shader specific code paths.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This patch fixes use of Altivec pack intrinsics on little-endian PowerPC
systems. Since little-endian operation only affects the load and store
instructions, the semantics of pack (and other) instructions that take
two input vectors implicitly change: the pack instructions still fill
a register placing values from the first operand into the "high" parts
of the register, and values from the second operand into the "low" parts
of the register, but since vector loads and stores perform an endian swap,
the high parts end up at high memory addresses.
To still achieve the desired effect, we have to swap the two inputs to
the pack instruction on little-endian systems. This is done automatically
by the back-end for instructions generated by LLVM, but needs to be done
manually when emitting intrisincs (which still result in that instruction
being emitted directly).
Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com>
Signed-off-by: Maarten Lankhorst <dev@mblankhorst.nl>
Instead we store a void pointer to the key, and cast it to
brw_wm_prog_key for fragment shader specific code paths.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Instead we store a brw_stage_prog_data pointer, and cast it to
brw_wm_prog_data for fragment shader specific code paths.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
BitSet::allocate() is being used with the expectation that it would
leave the bitfield untouched if its size hasn't changed, however,
the function always zeroed the last word, which led to obscure bugs
with live set computation.
This also fixes BitSet::resize(), which was broken, but luckily not
being used.
Cc: "10.2 10.3" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
This hasn't been enabled in a long time and is completely stale and
unnecessary. Remove, esp since it doesn't build.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Uniform values are in the UNIFORM register file, not the GRF register file.
Looking in virtual_grf_sizes makes no sense and only makes the output of
dump_instructions confusing.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
- include all headers in Makefile.sources
- bundle the scons buildscript, README and trace.xsl
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
- include all headers in Makefile.sources
- sort the list(s)
- bundle the android & scons buildscript
- include the headers' README & svga_dump.py
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
- include all headers in Makefile.sources
- sort the list(s)
- bundle the android & scons buildscript
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
- include all headers in Makefile.sources
- sort the list(s)
- bundle the android buildscript & README
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
- include all headers in Makefile.sources
- bundle the android buildscript
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
- include all headers in Makefile.sources
- sort the list(s)
- bundle the android buildscript & LLVM note
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
- include all headers in Makefile.sources
- sort the list(s)
- bundle the android buildscript & custom include
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
- include all headers in Makefile.sources
- sort the list(s)
- bundle the android buildscript & the tests
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
- include all headers in Makefile.sources
- sort the list(s)
- bundle the android buildscript
v2: Don't double-include the compiler sources.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
- include all headers in Makefile.sources
- bundle the scons buildscript
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
- include all headers in Makefile.sources
- sort the list(s)
- bundle the scons buildscript
v2: Don't double include the test sources.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
- include all headers in Makefile.sources
- sort the list(s)
- bundle the scons buildscript
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
- include all headers in Makefile.sources
- sort the list(s)
- bundle the scons buildscript
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
Commit 60d772cd9d1(st/vega: add headers and SConscript in
the tarball) meant to pick all the headers to be included in
the release tarball yet it missed a few.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
A few files were missing, namely:
- a few of the common headers
- the android + gdi sources
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
With the last revisions of commit 664c2d76947(gallium/ilo: cleanup
intel_winsys.h) we moved the header from winsys to drivers, but we
forgot to update the makefile.sources to reflect this.
Cc: Chia-I Wu <olvaffe@gmail.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
Currently, BLIT_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE* and
BLIT_MSAA_SHADER_2D_MULTISAMPLE_ARRAY_RESOLVE* shaders
in setup_glsl_msaa_blit_shader() are not recompiled
when the source buffer sample count changes. For example,
implementation continued using a 4X msaa shader, even if
source buffer changes from 4X msaa to 8x msaa. It causes
incorrect rendering.
This patch adds new enums in blit_msaa_shader, one for
each supported sample count, and uses them to store
msaa shaders.
Fixes following piglit tests on Broadwell:
ext_framebuffer_multisample-accuracy all_samples color
ext_framebuffer_multisample-accuracy all_samples depth_draw
ext_framebuffer_multisample-accuracy all_samples depth_resolve
ext_framebuffer_multisample-accuracy all_samples stencil_draw
ext_framebuffer_multisample-accuracy all_samples stencil_resolve
ext_framebuffer_multisample-formats all_samples
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstarnd <jason@jlekstrand.net>
Make sure to check the presence of the module in order to pick the
correct libs flag and before feeding them to the compiler/linker.
Current libXvMC*, libvdpau* and libomx_mesa depends unconditionally
upon xcb, due to their usage of the aux/vl gallium module.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Make sure to check the presence of the module in order to pick the
correct libs flag and before feeding them to the compiler/linker.
Current libEGL depends conditionally (when building with x11 platform)
upon xcb.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Make sure to check the presence of the module in order to pick the
correct libs flag and before feeding them to the compiler/linker.
Current libGL depends conditionally (when building with dri3) upon
xcb 1.9.3 and unconditionally on ancient xcb functions -
xcb_generate_id and xcb_request_check amongst others.
v2: Use PKG_CHECK_EXISTS() when checking for dri3 xcb.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=80848
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
When a texture is wrapped in a texture view, we can't trust the format in
the miptree itself. This patch allows us to pass the format seperately
through blorp so we can proprerly handled wrapped textures.
It's worth noting here that we can use the miptree format directly for
depth/stencil formats because they cannot be reinterpreted by a texture
view.
Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
CC: "10.3" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Before commit 04895f5c we would only reswizzle dot product instructions
(since they wrote the same value into all channels, and we didn't have
to think about anything else). That commit extended reswizzling to cases
when the swizzle was single valued -- i.e., writing the same result into
all channels.
But allowing reswizzling of arbitrary things is actually really easy and
is even less code. (Why didn't we do this in the first place?!)
total instructions in shared programs: 4266079 -> 4261000 (-0.12%)
instructions in affected programs: 351933 -> 346854 (-1.44%)
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Despite the comment above the function claiming otherwise, the function
did not reswizzle sources, which would lead to bad code generation since
commit 04895f5c, which began claiming we could do such swizzling when we
could not.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=82932
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The 'start' instruction is always in the current block, except for the
case of shader time, which emits code in a pattern seen no where else.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Otherwise, the basic block start/end IPs don't get updated properly,
leading to a broken CFG. This usually results in the following
assertion failure:
brw_fs_live_variables.cpp:141:
void brw::fs_live_variables::setup_def_use():
Assertion `ip == block->start_ip' failed.
Fixes KWin, WebGL demos, and a score of Piglit tests on Sandybridge and
earlier hardware.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The dump() methods don't alter the CFG or basic blocks, so we should
mark them as const. This lets you call them even if you have a const
cfg_t - which is the case in certain portions of the code (such as live
interval handling).
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
As older versions of gnu ld did not support --dynamic-list check to see
if it is supported before using it. Non gnu linkers such the apple one
likely lack this option as well.
Fixes the build on OpenBSD which has binutils 2.15 and 2.17.
The --dynamic-list option seems to been have introduced sometime after
binutils 2.17 was released as it is present in 2.18.
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Fixes binary garbage in the compilation logs caused by
compat::string::c_str() not being null-terminated (which is a bug on
its own that will be fixed in another commit).
Reported-by: EdB <edb+mesa@sigluy.net>
Reverts
* "i965: Modify state upload to allow 2 different sets of state atoms."
8e27a4d2b3
* "i965: Modify dirty bit handling to support 2 pipelines."
373143ed91
* "i965: Create a macro for checking a dirty bit."
c5bdf9be1e
Conflicts:
src/mesa/drivers/dri/i965/brw_context.h
* "i965: Create a macro for setting all dirty bits."
6f56e1424d
Conflicts:
src/mesa/drivers/dri/i965/brw_blorp.cpp
src/mesa/drivers/dri/i965/brw_state_cache.c
src/mesa/drivers/dri/i965/brw_state_upload.c
* "i965: Create a macro for setting a dirty bit."
88e3d404da
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
We can't rely on the value from the assembler if relative addressing is
used. So instead use the max of declared-consts (which does not include
compiler immediates) and what we get from the assembler (which does).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
all_delayed will also be true if we didn't attempt to schedule anything
due to no more instructions using current addr/pred. We rely on coming
in to block_sched_undelayed() to detect and clean up when there are no
more uses of the current addr/pred, which isn't necessarily an error.
This fixes a regression introduced in b823abed.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
The split between these two didn't make much sense. I'm going to want the
chance to look at uniform contents in optimization passes, and the QPU
emit I think is going to end up rewriting the uniforms stream.
This common init routine can be used by constructors for multiple program
types.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Debugging a regression in discard support was just too full of duplicate
instructions, so I decided to remove them instead of re-analyzing each of
them as I dumped their outputs in simulation.
There were troubles with bools without using native integers
(st_glsl_to_tgsi seemed to think bool true was 1.0f sometimes, when as a
uniform it's stored as ~0), and since I've got native integers other than
divide, I might as well just support them.
Before, we had some special opcodes like CMP and SNE that emitted multiple
instructions. Now, we reduce those operations significantly, giving
optimization more to look at for reducing redundant operations.
The downside is that QOP_SF is pretty special -- we're going to have to
track it separately when we're doing instruction scheduling, and we want
to peephole it into the instruction generating the destination write in
most cases (and not allocate the destination reg, probably. Unless it's
used for some other purpose, as well).
A bool is 0 or ~0, and KILL_IF takes a float arg that's <0 for discard or
>= 0 for not. By negating it, we ended up doing a floating point subtract
of (0 - ~0), which ended up as an inf. To make this actually work, we
need to convert the bool to a float.
Reviewed-by: Brian Paul <brianp@vmware.com>
While similar in layout, the size of the SVGA3dSize type may be smaller than
the struct drm_vmw_size type that is part of the ioctl interface. The kernel
driver could accordingly overwrite a memory area following the size variable
on the stack. Typically that would be another local variable, causing
breakage in, for example, ubuntu 12.04.5 where the handle local variable
becomes overwritten.
v2: Fix whitespace errors
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Cc: "10.1 10.2 10.3" <mesa-stable@lists.freedesktop.org>
As explained in the previous commit, we want to avoid the possibility of
integer-multiplication overflow while allocating buffers.
In these two cases, the final allocation size is the product of three values:
one variable and two that are fixed constants at compile time.
In this commit, we move the explicit multiplication to involve only the
compile-time constants, preventing any overflow from that multiplication, (and
allowing calloc to catch any potential overflow from the remainining implicit
multiplication).
Reviewed-by: Matt Turner <mattst88@gmail.com>
In commit 32f2fd1c5d, several calls to
_mesa_calloc(x) were replaced with calls to calloc(1, x). This is strictly
equivalent to what the code was doing previously.
But for cases where "x" involves multiplication, now that we are explicitly
using the two-argument calloc, we can do one step better and replace:
calloc(1, A * B);
with:
calloc(A, B);
The advantage of the latter is that calloc will detect any overflow that would
have resulted from the multiplication and will fail the allocation, (whereas
the former would return a small allocation). So this fix can change
potentially exploitable buffer overruns into segmentation faults.
Reviewed-by: Matt Turner <mattst88@gmail.com>
It's been altering the tree and reporting "false" since January 2011.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Previously, opt_copy_propagation_elements would always rewrite the
instruction stream, even if was the same thing as before. In order to
report progress correctly, we'll need to bail if the suggested
replacement is identical (or equivalent) to the original code.
This also introduced unnecessary noop swizzles, as far as I can tell.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Previously, if chans < 4, we passed uninitialized stack garbage to the
ir_swizzle constructor for the excess components. Thankfully, it
ignores that data, as it's unnecessary, so no harm actually comes of it.
However, it's obviously better to initialize it.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
ir_triop_csel can return a boolean expression, so we need to handle it
here; we simply forgot when we added it.
Fixes Piglit's EXT_shader_integer_mix/{vs,fs}-mix-if-bool.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Cc: mesa-stable@lists.freedesktop.org
All shader stages have these fields, so it makes sense to store them in
the common base structure, rather than duplicating them in each.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
In commit 46d03d37bf I renamed a Makefile target
from md5 to checksums, (as we switched from MD5 checksums to SHA-256
checksums, so the more general name is more future proof).
But that commit missed one mention of "md5" as a dependency of the .PHONY
target. Rename that here as well.
According to the GLSL 1.40 spec, section 5.7 Structure and Array Operations:
"Array elements are accessed using an expression whose type is int or uint."
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
The default case was accidentally clearing RADEON_FLAG_CPU_ACCESS from the
previous fall-through cases.
Reported-by: Mathias Fröhlich <Mathias.Froehlich@gmx.net>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Coverity pointed out we never dropped the lock here, so fix
it by using a common exit path.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
tex-miplevel-selection was hammering my memory manager with primconverts
on individual quads. This gets all those converted IBs packed into larger
IBs.
Reviewed-by: Rob Clark <robclark@freedesktop.org>
This is part of fixing extremely long runtimes on some piglit tests that
involve streaming vertex reuploads due to format conversions, and will
similarly be important for X performance, which relies on these flags.
A meta begin/end pair with MESA_META_DRAW_BUFFERS will change visible GL
state. We recreate the draw buffer enums from the buffer bitfield, which
changes GL_BACK to GL_BACK_LEFT (and GL_FRONT to GL_FRONT_LEFT).
This commit modifes the save/restore logic to instead copy the buffer enums
from the gl_framebuffer and then set them on restore using
_mesa_drawbuffers().
It's not clear how this breaks the benchmark in 82796, but fixing meta to not
leak the state change fixes the regression.
No piglit regressions.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=82796
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Cc: mesa-stable@lists.freedesktop.org
Coverity reported this, and I think this is the right solution,
since cache->items is struct cache_item ** not struct cache_item *,
we also realloc it using struct cache_item * at some point.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Not all of these are used in every context, so this can make a
significant difference for short-lived contexts such as in piglit tests.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
If we fails in reserve_explicit_locations, we leak uniform_map.
Reported-by: coverity scanner.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
i965 will have more than 32 bits when BRW_STATE_COMPUTE_PROGRAM is added.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
The set of state atoms for compute shaders is currently empty; it will
be filled in by future patches.
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
The hardware state for compute shaders is almost entirely orthogonal
to the hardware state for 3D rendering. To avoid sending unnecessary
state to the hardware, we'll need to have a separate set of state
atoms for the compute pipeline and the 3D pipeline. That means we
need to maintain two separate sets of dirty bits to determine which
state atoms need to be run.
But the dirty bits are not completely independent; for example, if
BRW_NEW_SURFACES is flagged while doing 3D rendering, then not only do
we need to re-run 3D state atoms that depend on BRW_NEW_SURFACES, but
we also need to re-run compute state atoms that depend on
BRW_NEW_SURFACES. But we'll also need to re-run those state atoms the
next time the compute pipeline is run.
To accomplish this, we record two sets of dirty bits, one for each
pipeline. When bits are dirtied (via SET_DIRTY_BIT() or
SET_DIRTY_ALL()) we set them to the dirty state in both pipelines.
When brw_state_upload() is run, we clear the dirty bits just for the
pipeline that was run.
Note that since the number of pipelines is known at compile time to be
2, the compiler should unroll the loops in SET_DIRTY_BIT() and
SET_DIRTY_ALL().
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
This one path doesn't goto fail, so it seems to leak dec.
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Dave Airlie <airlied@redhat.com>
With VP2, nv50_miptree is faked because the underlying bo's have to be
laid out in a certain way. This is done by adjusting the address. Make
sure that blits (and everything else for consistency) use the mt address
rather than the bo address as a base.
This fixes retrieving chroma plane with VDPAU.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=82255
Tested-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.2 10.3" <mesa-stable@lists.freedesktop.org>
When constant folding a MAD operation, we first fold the multiply and
generate an ADD. However we do so without making sure that the immediate
can be handled in the saturate case. If it can't, load the immediate in
a separate instruction.
Reported-by: Tiziano Bacocco <tizbac2@gmail.com>
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.2 10.3" <mesa-stable@lists.freedesktop.org>
Experimentally, the sampler doesn't appear to like these, neither as
buffer nor as rect textures. So remove 1D from the list of texture types
to make linear when used for staging.
This fixes the OSD in mplayer for VDPAU.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.2 10.3" <mesa-stable@lists.freedesktop.org>
Samplers are only defined up to num_samplers, so set all samplers above
nr to NULL so that we don't try to read them again later.
Tested-by: Christian Ruppert <idl0r@qasl.de>
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.2 10.3" <mesa-stable@lists.freedesktop.org>
We always left them enabled, which turned off HiZ in some cases.
This should improve performace with Hyper-Z.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
This should be as fast as no HTILE for stencil. I think we can still get full
performance with depth-only rendering even if stencil is present in the buffer
but not used, but I'm not 100% sure. This may be revisited when HiS and fast
stencil clear are implemented.
This fixes a hang in Brutal Legend.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=64471
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
This is a golden setting on RV740, but there is a hw bug which recommends
setting it on all R7xx chipsets.
Acked-by: Michel Dänzer <michel.daenzer@amd.com>
It's almost the same.
This enables tiling for HTILE. It also enables Hyper-Z for other texture
targets (1D, 1D_ARRAY, 2D_ARRAY, CUBE, CUBE_ARRAY, 3D, RECT).
2D array depth textures are tested by Unigine Sanctuary and my new piglit
test.
Acked-by: Michel Dänzer <michel.daenzer@amd.com>
This should make a machine which is running piglit more responsive at times.
e.g. streaming-texture-leak can easily eat 600 MB because of how fast it
creates new textures.
We were only using it to get at its type, which we already know because
it's a builtin variable.
Signed-off-by: Connor Abbott <connor.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
We were only using it to get at its type, which we already know because
it's a builtin variable.
v2 (Ken): Rebase on Matt's optimized gl_FrontFacing calculations.
Signed-off-by: Connor Abbott <connor.abbott@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The replicated data clear shader needs to be SIMD16, or else the GPU
will hang. So, compile it even if INTEL_DEBUG=no16 is set.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Current method of generating distribution tar-balls involves manually
invoking make + target name in the appropriate places. This temporary
solution is used until we get 'make dist' working.
Currently it does not work, as in order to have the target (which is
also a filename) available in the final Makefile we need to add a PHONY
target + use the correct target name.
Cc: "10.2 10.3" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
v3: Since the fs backend can emit saturate as a separate instruction, there is
no need to detect for min/max instructions and to rewrite the instruction tree
accordingly. On the other hand, we don't need to emit a separate saturated
mov either when the expression generating src can do saturate directly.
v4: Add can_do_saturate() check before enabling saturate modifer (Ken)
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
When sel conditon is bounded within 0 and 1.0. This allows code as:
mov.sat a b
sel.ge dst a 0.25F
To be propagated as:
sel.ge.sat dst b 0.25F
v3: - Syntax clarifications in inst->saturate assignment
- Remove extra parenthesis when assigning src_reg value
from copy_entry (Matt Turner)
v4: - Take channels into consideration when propagating saturated instructions.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
When sel conditon is bounded within 0 and 1.0. This allows code as:
mov.sat a b
sel.ge dst a 0.25F
To be propagated as:
sel.ge.sat dst b 0.25F
v3: Syntax clarifications in inst->saturate assignment (Matt Turner)
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
v2: - Output max(saturate(x),b) instead of saturate(max(x,b))
- Make sure we do component-wise comparison for vectors (Ian Romanick)
v3: - Add missing condition where the outer constant value is > 0.0 and
inner constant is 1.0.
- Fix comments to show that the optimization is a commutative operation
(Matt Turner)
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
v2: - Output min(saturate(x),b) instead of saturate(min(x,b)) suggested by Ilia Mirkin
- Make sure we do component-wise comparison for vectors (Ian Romanick)
v3: - Add missing condition where the outer constant value is zero and
inner constant is < 1
- Fix comments to reflect we are doing a commutative operation (Matt Turner)
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
v2: - Check that the base type is float (Ian Romanick)
v3: - Make sure comments reflect that we are doing a commutative operation
- Add missing condition where the inner constant is 1.0 and outer constant is 0.0
- Make indexing of operands easier to read (Matt Turner)
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
Suggested by Matt. This patch combines and moves back the code-generation
functions from generate_vec4_instruction() into generate_code(). Makes
generate_code() a bit larger, but helps us to count loops in a
straightforward manner.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Abdiel Janulgue <abdiel.janulgue@linux.intel.com>
brw_meta_fast_clear.c:211:17: warning: 'x_scaledown' may be used
uninitialized in this function [-Wmaybe-uninitialized]
unsigned int x_scaledown, y_scaledown;
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
There are some cases where the scheduler can get itself into impossible
situations, by scheduling the wrong write to pred or addr register
first. (Ie. it could end up being unable to schedule any instruction if
some instruction which depends on the current addr/reg value also
depends on another addr/reg value.)
To solve this we'd need to be able to insert extra mov instructions
(which would also help when register assignment gets into impossible
situations). To do that, we'd need to move the nop padding from sched
into legalize.
But to start with, just detect when we get into an impossible situation
and bail, rather than sitting forever in an infinite loop. This way it
will at least fall back to the old compiler, which might even work if
you are lucky.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
All of the GL image enums fit in 16-bits.
Also move the fields from the anonymous "image" structucture to the next
higher structure. This will enable packing the bits with the other
bitfield.
Valgrind massif results for a trimmed apitrace of dota2:
n time(i) total(B) useful-heap(B) extra-heap(B) stacks(B)
Before (32-bit): 76 40,572,916,873 68,831,248 63,328,783 5,502,465 0
After (32-bit): 70 40,577,421,777 68,487,584 62,973,695 5,513,889 0
Before (64-bit): 60 36,822,640,058 96,526,824 88,735,296 7,791,528 0
After (64-bit): 74 37,124,603,758 95,891,808 88,466,712 7,425,096 0
A real savings of 346KiB on 32-bit and 262KiB on 64-bit.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The only values allowed are 0 and 1, and the value is checked before
assigning.
Valgrind massif results for a trimmed apitrace of dota2:
n time(i) total(B) useful-heap(B) extra-heap(B) stacks(B)
Before (32-bit): 74 40,580,119,657 69,186,544 63,506,327 5,680,217 0
After (32-bit): 76 40,572,916,873 68,831,248 63,328,783 5,502,465 0
Before (64-bit): 89 36,822,971,897 96,526,616 88,735,296 7,791,320 0
After (64-bit): 60 36,822,640,058 96,526,824 88,735,296 7,791,528 0
A real savings of 173KiB on 32-bit and no change on 64-bit.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Just use ir_variable::data.binding... because that's the where the
binding is stored for everything else that can use layout(binding=).
Valgrind massif results for a trimmed apitrace of dota2:
n time(i) total(B) useful-heap(B) extra-heap(B) stacks(B)
Before (32-bit): 50 40,564,927,443 69,185,408 63,683,871 5,501,537 0
After (32-bit): 74 40,580,119,657 69,186,544 63,506,327 5,680,217 0
Before (64-bit): 59 36,822,048,449 96,526,888 89,113,000 7,413,888 0
After (64-bit): 89 36,822,971,897 96,526,616 88,735,296 7,791,320 0
A real savings of 173KiB on 32-bit and 368KiB on 64-bit.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
The VertexProgram and FragmentProgram have a Cache member for dealing
with fixed function programs. There are no fixed function geometry
programs, so this should never have existed, and was just copy and
pasted.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
I actually screwed that up in 754319490f,
mistakenly thinking the code actually wanted the non-nan result before.
So, introduce that missing nan behavior case and use that instead.
For sse, there's no actual change in the resulting code at all, the fallback
code wouldn't have done the right thing though.
Of course, the actual issue I saw with pow() was completely unrelated...
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Pretty trivial, just fill in the offsets and such. The implementation
is near 100% copy and paste from llvmpipe. Should be useful for debugging.
No piglit change when not using SOFTPIPE_USE_LLVM=1.
Now that it can do the same tests with and without using llvm for vs/gs,
with llvm more pass, the only things failing only with llvm seems to be
edgeflags tests and vs/gs-pow-float-float (and for the latter I'm not
convinced the zero tolerance it requires is somehow mandated by glsl).
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
The code is all in place now so enable it.
Seems to pass all relevant piglit tests (just like cube maps, some of the
cube map array tests need GALLIVM_DEBUG=no_quad_lod,no_rho_approx)
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Pretty easy, just make sure that all paths testing for PIPE_TEXTURE_CUBE
also recognize PIPE_TEXTURE_CUBE_ARRAY, and add the layer * 6 calculation
to the calculated face.
Also handle it for texture size query, looks like OpenGL wants the number
of cubes, not layers (so need division by 6).
No piglit regressions.
v2: fix up adding cube layer to face for seamless filtering (needs to happen
after calculating per-sample face). Undetected by piglit unfortunately.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com> (v1)
Not sure why it was there but it is definitely not an error if gs outputs are
infs/nans. Besides, the outputs can be ints, in which case any small negative
number asserted.
This fixes piglit's texelFetch gs isamplerXX crashes with softpipe (down from
14 to 2).
Bug https://bugs.freedesktop.org/show_bug.cgi?id=80012
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
piglit tex-miplevel-selection nowadays doesn't use repeat wrap mode due to
sampler objects any longer, however at the time of the clear the wrap mode
is still illegal and at this point we get to verify the state, including
samplers (even though they won't get used), and because mesa doesn't treat
it as an incomplete texture as the spec says it should, we hit the assertion.
Just warn about this for now instead.
Gets crashes down from 44 to 14 in a piglit run (all were in various tests of
tex-miplevel-selection with texture rectangles). Though just about all
tex-miplevel-selection tests fail anyway for other reasons.
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Just handle as ordinary 2d / 2d array resources. Prevents an assertion failure
with softpipe and piglit glsl-resource-not-bound 2DMS/2DMSArray tests.
While here also fix TXD shadowCube similarly, which fixes the crash with piglit
tex-miplevel-selection textureGrad CubeShadow (the test will still fail due to
softpipe being broken).
This fixes https://bugs.freedesktop.org/show_bug.cgi?id=80011
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
This was meant for softpipe to not crash at some point if vertex texturing
was used. It is, however, fishy because it uses values from
draw_set_samplers/draw_set_sampler_views and not from the shader key. Albeit
we should still in all cases actually generate a new shader if this changes
(because the samplers and views themselves are in the key) I don't want to
think again wondering if that's really correct in the future.
Besides, at least today, it does not actually work for softpipe, as this was
relying on softpipe not actually calling draw_set_samplers/sampler_views at
all - I've verified it crashes regardless (if there were a tex instruction in
the vs, which normally should not happen anyway). For drivers which do indeed
not call these functions because they don't support vertex texturing at all
(r300), this should still not crash because the static texture data is all
zero, which causes the sampling functions to take an early out (same as is done
if no texture is bound at the slot used for sampling - verified with hacked up
softpipe).
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
mesa was creating a cube map array texture with just one layer, which is
not legal. This caused an assertion failure when using that texture later
in llvmpipe (when enabling cube map arrays) since it verifies the number
of layers in the view is divisible by 6 (the sampling code might well crash
randomly otherwise) with piglit glsl-resource-not-bound CubeArray -fbo -auto.
v2: use appropriately sized texel array...
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com> (v1)
The source can be a register as well as an immediate, and disassembling
a register as an immediate can have some strange results.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
It is meant to be private within the actual winsys. Remove it from
the exported header, and fold it into it's only user.
Cc: Alexander von Gluck IV <kallisti5@unixzen.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
This define is always set and it had no real purpose according to
git log. Seems like it is a leftover from the vl/vdpau prototype
stage.
Cc: Christian König <christian.koenig@amd.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Namely
- the private header (xa_priv.h)
- README and
- xa-indent
Sort the sources list while we're here.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
This reverts commit 6a19bb56e0.
The above commit disabled the default build of xvmc as the xvmc tests
were failing. As pointed out by Ilia, the tests are "broken by design"
as they do not test the object that is build but the one that is
installed and setup on the workstation.
With previous commit we moved the programs from the 'make check' to
noinst automake target. This way they won't be run but will be around
for people to use them.
Cc: Tom Stellard <thomas.stellard@amd.com>
All the tests require an installed and setup XvMC, thus they
are not good candidates for 'make check'.
Keep them around as the user might want to actually test the
implementation post installation/setup.
Cc: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: Tom Stellard <thomas.stellard@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Add the final remaining files into the tarball (make dist), namely:
- SConscripts
- Non-autotooled winsys' - android, gdi and hgl.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
- Include the headers within.
- Update scons to use them.
- Drop useless include (gallium/drivers) from scons.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
- Drop duplicate include compiler directives.
- Leave the sw/ prefix for all the software winsys headers.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
- Add top_srcdir/src/gallium/winsys to GALLIUM_DRIVER_C{XXFLAGS}.
- Remove top_srcdir/src/gallium/drivers/radeon from the includes.
As a result:
- Common radeon headers are prefixed with 'radeon/'
- Winsys header inclusion is prefixed 'radeon/drm'
Cc: Marek Olšák <marek.olsak@amd.com>
Cc: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Make the header location, inclusion and contents more common with
its i915,r* and nouveau counterparts:
- Move the header within drivers/ilo.
- Separate out intel_winsys_create_for_fd into 'drm_public' header.
- Cleanup the compiler includes.
v2: Move the header to drivers/ilo. Suggested by Chia-I.
v3: Correct intel_winsys.h inclusion. Spotted by Chia-I.
Cc: Chia-I Wu <olvaffe@gmail.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
V2: moved test for the VertexAttrib*Pointer() functions
to update_array(), and made constant available for drivers to set
Signed-off-by: Timothy Arceri <t_arceri@yahoo.com.au>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
The base instance needs to be passed to the jited function, otherwise the
instanced data fetch will only work with the same start instance when the
jit function was created (and baking that into the key instead is not a viable
option).
This fixes piglit arb_base_instance-drawarrays (modulo some unrelated
core/compat context trouble I get for the test).
And fix the pipe cap bit in llvmpipe for it now that it actually works (it
already worked for softpipe).
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
The docs were never really up to date for them, missing just about everything.
So mark them off as all done for GL 3.3 (though softpipe is in fact quite
broken for some newer things especially wrt texturing, and both don't have
compliant, real msaa support). And add the extensions missing too (no
guarantee of completeness).
Reviewed-by: Dave Airlie <airlied@gmail.com>
* IEEE Std 1003.1-2001 placed strcasecmp() in strings.h.
* ISO C99 doesn't mention strcase* in string.h
* On all platforms I could find, strcasecmp is in strings.h and string.h
as a compatibility layer for software written pre-2001 POSIX
* Technically strcasecmp should be only in strings.h and the man
pages back this up.
* Tested build on CentOS and Haiku
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This is never used. There is another token "OUTPUT" which the lexer can
generate, though. This has been around since the dawn of time; is most
likely a typo.
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Now that qpu_inst() ignores the WADDR from the other half of the
instruction, we can set both the ADD and MUL WADDRs in the NOP helper.
Thanks to that, we also no longer need to qpu_inst(NOP, NOP).
This allows setting the opposite-side WADDR to NOP (a non-zero value) in
qpu_* helpers, so that we don't need to qpu_inst() merge them with NOPs
all the time just to get the waddr set.
Fixes piglit draw-vertices and gl-2.0-vertexattribpointer on vc4, where
I'm only advertising R32F to RGBA32F support so far.
Note: regresses gl-1.5-normal3b3s-invariance due to introduced flushes and
missing depth buffer load/store support in the driver.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Individual caps made supporting new fallbacks more complicated than it
needed to be. Instead, just make a table of fallbacks at context init
time.
v2: Fix inverted "do we need to install vbuf?" flagging caught by Marek.
Reviewed-by: Marek Olšák <marek.olsak@amd.com> (v2)
When I made the shader cache take the .fs member and moved the binding
point to .bind_fs, I failed to update these. Fixes crashes in
copyteximage-related tests.
The patch fixes the build on Oracle Solaris.
CC os/os_misc.lo
"os/os_misc.c", line 59: #error: unexpected platform in os_sysinfo.c
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
We've found that there's a buffer overrun bug in flex that's triggered by
using alternation in a lookahead pattern.
Fortunately, we don't need to match the exact {NEWLINE} expression to
detect an empty pragma. It suffices to verify that there are no non-space
characters before any newline character. So we can use a simple [\r\n] to
get the desired behavior while avoiding the flex bug.
Fixes the regression of piglit's 17000-consecutive-chars-identifier test,
(which has been crashing since commit
04e40fd337 ).
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=82472
Signed-off-by: Carl Worth <cworth@cworth.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
CC: <mesa-stable@lists.freedesktop.org>
The optimization relies on CMP setting the destination to 0, which is
equivalent to 0.0f. However, early platforms only set the least
significant byte, leaving the other bits undefined. So, we must disable
the optimization on those platforms.
Oddly, Sandybridge wasn't reported as broken. The PRM states that it
only sets the LSB, but the internal documentation says that it follows
the IVB behavior. Since it wasn't reported as broken, we believe it
really does follow the IVB behavior.
v2: Allow the optimization on Sandybridge (requested by Matt).
+32 piglits on Ironlake.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?=79963
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Operating on this code,
B0: ...
cmp.ne.f0(8)
(+f0) if(8)
B1: break(8)
B2: endif(8)
We can delete B2 without attempting to merge any blocks, since the
break/continue instruction necessarily ends the previous block.
After deleting the if instruction, we attempt to merge blocks B0 and B1.
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
This pass deletes an IF/ELSE/ENDIF or IF/ENDIF sequence, or the ELSE in
an ELSE/ENDIF sequence.
In the typical case (where IF and ENDIF) aren't the only instructions in
their basic blocks, we can simply remove the instructions (implicitly
deleting the block containing only the ELSE), and attempt to merge
blocks B0 and B2 together.
B0: ...
(+f0) if(8)
B1: else(8)
B2: endif(8)
...
If the IF or ENDIF instructions are the only instructions in their
respective basic blocks (which are deleted by the removal of the
instructions), we'll want to instead merge the next blocks.
Both B0 and B2 are possibly removed by the removal of if & endif.
Same situation for if/endif. E.g., in the following example we'd remove
blocks B1 and B2, and then attempt to combine B0 and B3.
B0: ...
B1: (+f0) if(8)
B2: endif(8)
B3: ...
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
... rather than pointing directly to the associated instruction. This
will let us set the block containing the IF statement's else-pointer to
NULL, when we delete a useless ELSE instruction, as in the case
(+f0) if(8)
...
else(8)
endif(8)
Also, remove the pointer to the ENDIF, since it's unused, and it was
also potentially wrong, in the case of a basic block containing both an
ENDIF and an IF instruction:
endif(8)
cmp.ne.f0(8) ...
(+f0) if(8)
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
To avoid invalidating and recreating the control flow graph. Also stop
invalidating the CFG in places we didn't add or remove an instruction.
cfg calculations: 202951 -> 80307 (-60.43%)
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Will let us avoid invalidating the CFG if the optimization pass has
removed instructions using the new basic block methods.
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Fixes piglit glsl-fs-discard-01 and -03, and allows a lot of mesa demos to
start running. glsl-fs-discard-02 has a problem where the first tile is
not getting stored on the first render.
This should improve performance on real hardware by allowing more shader
instances to run in parallel. It also fixes assertion failures in tests
that don't emit a fragment color, since otherwise we didn't have enough
instructions to fit our signals in.
The spec citation talked about A and B, and I proceeded to pay no
attention to whether the waddrs were for A or B. As a result, this pair
of instructions would claim to conflict:
mov ra4, ra4 ; nop nop, r0, r0
mov.ns ra4, rb4 ; nop nop, r0, r0
Now that tiling is in place, we can expose the other formats. Depth is
still broken (need to make changes in the shader), but if you don't expose
it things crash all over. SNORM is dropped, but we could re-add it later
with some shader fixes to handle converting between [0,1] and [-1,1].
This still treats everything as RGBA8888 for the most part, same as
before. This is a prerequisite for handling other texture formats, since
only RGBA8888 has a raster-layout mode.
It meant that LUMALPHA was being marked as *many* miplevels, and
unsurprisingly wouldn't validate. On the other hand, some miplevel counts
wouldn't get the small mips validated at all.
The header is used by DRI1 drivers, which we've removed a while
back. Now only the dri1 loader in libGL is using it, so let's
move it in src/glx, and prefix it accordingly.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
This flag was set to true for the atomic counter intrinsics, but it
never got plumbed through the linker, so by the time it got to the
backends it would always be set to the false. The current i965 backend
code doesn't use is_intrinsic, so this should not change any existing
code, but it's useful for codepaths that want to distinguish between
intrinsics and non-intrinsics without using strcmp.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Connor Abbott <connor.abbott@intel.com>
This captures all of the steps I have been following in making releases for
the past year or so. This way, the instructions should be sound for anyone who
would like to take over the release process going forward.
This change will double cache size for branches which have a lower
LP_MAX_SHADER_VARIANTS limit (it will not do anything on master).
The reason is that nowadays shaders tend to be quite a bit larger than they
were (they were big when llvmpipe didn't have a fs loop, got much smaller with
that loop, and since then have gradually increased quite a bit though still
smaller than without the fs loop for various reasons - among them being d3d10
compliance, usage of 8-wide vectors, non-swizzled blend code). Thus effectively
less shaders would be cached (unless they were very small and the variant limit
was hit first). Also, since we're getting rid of the IR nowadays, the cached
shaders shouldn't need all that much memory actually.
This captures the set of rules I have been using for stable-branch management,
(starting with a discussion on the mesa-dev mailing list on July 2013, and
then refined through my own experience of performing stable-branch releases
since then).
We switched to these several stable releases ago, (since the MD5 algorithm has
been broken for some time), but only now did I get around to fixing this in
the Makefile rather than just performing this step manually.
CC: "10.2 10.3" <mesa-stable@lists.freedesktop.org>
v2:
- Move dri*_query_renderer_* into their respective dri*_priv.h headers
- Drop then unnneeded include of dri2.h from dri2_query_renderer.c
- Rename dri2_query_renderer.c as dri_common_query_renderer.c, as it's contents
now are used for more than dri[23]
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82709">Bug 82709</a> - OpenCL not working on radeon hainan</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82814">Bug 82814</a> - glDrawBuffers(0, NULL) segfaults in _mesa_drawbuffers</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=83079">Bug 83079</a> - [NVC0] Dota 2 (Linux native and Wine) crash with Nouveau Drivers</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=83355">Bug 83355</a> - FTBFS: src/mesa/program/program_lexer.l:122:64: error: unknown type name 'YYSTYPE'</li>
</ul>
<h2>Changes</h2>
<p>Adam Jackson (1):</p>
<ul>
<li>radeonsi: Don't use anonymous struct trick in atom tracking</li>
</ul>
<p>Alex Deucher (2):</p>
<ul>
<li>radeonsi: add new CIK pci ids</li>
<li>radeonsi: add new SI pci ids</li>
</ul>
<p>Andreas Boll (1):</p>
<ul>
<li>winsys/radeon: fix nop packet padding for hawaii</li>
</ul>
<p>Anuj Phogat (1):</p>
<ul>
<li>i965: Bail on vec4 copy propagation for scratch writes with source modifiers</li>
</ul>
<p>Brian Paul (1):</p>
<ul>
<li>mesa: fix NULL pointer deref bug in _mesa_drawbuffers()</li>
</ul>
<p>Carl Worth (2):</p>
<ul>
<li>docs: Add sha256 sums for the 10.2.6 release</li>
<li>Makefile: Switch from md5sums to sha256sums</li>
</ul>
<p>Dave Airlie (1):</p>
<ul>
<li>i965: add missing parens in vec4 visitor</li>
</ul>
<p>Emil Velikov (17):</p>
<ul>
<li>configure.ac: bail out if building gallium_gbm without gallium_egl</li>
<li>android: gallium/nouveau: fix include folders, link against libstlport</li>
<li>android: egl/main: fixup the nouveau build</li>
<li>automake: gallium/freedreno: drop spurious include dirs</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=77493">Bug 77493</a> - lp_test_arit fails with llvm >= llvm-3.5svn r206094</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82539">Bug 82539</a> - vmw_screen_dri.lo In file included from vmw_screen_dri.c:41: vmwgfx_drm.h:32:17: error: drm.h: No such file or directory</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=84140">Bug 84140</a> - mplayer crashes playing some files using vdpau output</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=84662">Bug 84662</a> - Long pauses with Unreal demo Elemental on R9270X since : Always flush the HDP cache before submitting a CS to the GPU</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=85267">Bug 85267</a> - vlc crashes with vdpau (Radeon 3850HD) [r600]</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=74863">Bug 74863</a> - [r600g] HyperZ broken on RV770 and CYPRESS (Left 4 Dead 2 trees corruption) bisected!</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=77865">Bug 77865</a> - [BDW] Many Ogles3conform framebuffer_blit cases fail</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=78225">Bug 78225</a> - Compile error due to undefined reference to `gbm_dri_backend', fix attached</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=78258">Bug 78258</a> - make check link_varyings.gl_ClipDistance failure</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=78403">Bug 78403</a> - query_renderer_implementation_unittest.cpp:144:4: error: expected primary-expression before ‘.’ token</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=78468">Bug 78468</a> - Compiling of shader gets stuck in infinite loop</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=78537">Bug 78537</a> - no anisotropic filtering in a native Half-Life 2</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=78581">Bug 78581</a> - OpenCL: clBuildProgram prints error messages directly rather than storing them</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=78648">Bug 78648</a> - Texture artifacts in Kerbal Space Program</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=78665">Bug 78665</a> - macros in builtin_functions.cpp make invalid assumptions about M_PI definitions</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=78692">Bug 78692</a> - Football Manager 2014, gameplay rendered black & white</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=78716">Bug 78716</a> - Fix Mesa bugs for running Unreal Engine 4.1 Cave effects demo compiled for Linux</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=78803">Bug 78803</a> - gallivm/lp_bld_debug.cpp:42:28: fatal error: llvm/IR/Module.h: No such file or directory</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=78888">Bug 78888</a> - test_eu_compact.c:54:3: error: implicit declaration of function ‘brw_disasm’ [-Werror=implicit-function-declaration]</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79029">Bug 79029</a> - INTEL_DEBUG=shader_time is full of lies</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79095">Bug 79095</a> - x86/common_x86.c:348:14: error: use of undeclared identifier 'bit_SSE4_1'</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79263">Bug 79263</a> - Linking error in egl_gallium.la when compiling 32 bit on multiarch</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79294">Bug 79294</a> - Xlib-based build broken on non x86/x86-64 architectures</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79373">Bug 79373</a> - Non-const initializers for matrix and vector constructors</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79382">Bug 79382</a> - build error: multiple definition of `loader_get_pci_id_for_fd'</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79616">Bug 79616</a> - L4D2 crash on startup</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79724">Bug 79724</a> - switch statement type check</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79729">Bug 79729</a> - [i965] glClear on a multisample texture doesn't work</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79809">Bug 79809</a> - radeonsi: mouse cursor corruption using weston on AMD Kaveri</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79823">Bug 79823</a> - [NV30/gallium] Mozilla apps freeze on startup with nouveau-dri-10.2.1 libs on dual-screen</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79885">Bug 79885</a> - commit b52a530 (gallium/egl: st_profiles are build time decision, treat them as such) broke egl</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79903">Bug 79903</a> - [HSW Bisected]Some Piglit and Ogles2conform cases fail</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=81020">Bug 81020</a> - [radeonsi][regresssion] Wireframe of background rendered through objects in Half-Life 2: Episode 2 with MSAA enabled</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82159">Bug 82159</a> - No rule to make target `../../../../src/mesa/libmesa.la', needed by `collision'.</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82255">Bug 82255</a> - [VP2] Chroma planes are vertically stretched during VDPAU playback</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82268">Bug 82268</a> - Add support for the OpenRISC architecture (or1k)</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82428">Bug 82428</a> - [radeonsi,R9 270X] System lockup when using mplayer/mpv with VDPAU</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82483">Bug 82483</a> - format_srgb.h:145: undefined reference to `util_format_srgb_to_linear_8unorm_table'</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82517">Bug 82517</a> - [RADEONSI,VDPAU] SIGSEGV in map_msg_fb_buf called from ruvd_destroy, when closing a Tab with accelerated video player</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82534">Bug 82534</a> - src\egl\main\eglapi.h : fatal error LNK1107: invalid or corrupt file: cannot read at 0x2E02</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82536">Bug 82536</a> - u_current.h:72: undefined reference to `__imp__glapi_Dispatch'</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82538">Bug 82538</a> - Super Maryo Chronicles fails with st/mesa assertion failure</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82539">Bug 82539</a> - vmw_screen_dri.lo In file included from vmw_screen_dri.c:41: vmwgfx_drm.h:32:17: error: drm.h: No such file or directory</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=83355">Bug 83355</a> - FTBFS: src/mesa/program/program_lexer.l:122:64: error: unknown type name 'YYSTYPE'</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=70410">Bug 70410</a> - egl-static/Makefile: linking fails with llvm >= 3.4</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=72685">Bug 72685</a> - [radeonsi hyperz] Artifacts in Unigine Sanctuary</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=72819">Bug 72819</a> - [855GM] Incorrect drop shadow color on windows and strange white rectangle when showing/hiding GLX-dock...</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=74563">Bug 74563</a> - Surfaceless contexts are not properly released by DRI drivers</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=74863">Bug 74863</a> - [r600g] HyperZ broken on RV770 and CYPRESS (Left 4 Dead 2 trees corruption) bisected!</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=75011">Bug 75011</a> - [hyperz] Performance drop since git-01e6371 (disable hyperz by default) with radeonsi</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=75112">Bug 75112</a> - Meta Bug for HyperZ issues on r600g and radeonsi</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=76252">Bug 76252</a> - Dynamic loading/unloading of opengl32.dll results in a deadlock</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=76861">Bug 76861</a> - mid3 generates slow code for constant arguments</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=77957">Bug 77957</a> - Variably-indexed constant arrays result in terrible shader code</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=78468">Bug 78468</a> - Compiling of shader gets stuck in infinite loop</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79155">Bug 79155</a> - [Tesseract Game] Global Illumination: Medium Causes Color Distortion</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=79462">Bug 79462</a> - [NVC0/Codegen] Shader compilation falis in spill logic</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=80050">Bug 80050</a> - [855GM] Incorrect drop shadow color under windows in Cinnamon persists with MESA 10.1.</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=80247">Bug 80247</a> - Khronos conformance test ES3-CTS.gtf.GL3Tests.transform_feedback.transform_feedback_vertex_id fails</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=80561">Bug 80561</a> - Incorrect implementation of some VDPAU APIs.</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82538">Bug 82538</a> - Super Maryo Chronicles fails with st/mesa assertion failure</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82539">Bug 82539</a> - vmw_screen_dri.lo In file included from vmw_screen_dri.c:41: vmwgfx_drm.h:32:17: error: drm.h: No such file or directory</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=82796">Bug 82796</a> - [IVB/BYT-M/HSW/BDW Bisected]Synmark2_v6.0_OglTerrainFlyInst/OglTerrainPanInst cannot run as image validation failed</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=83148">Bug 83148</a> - Unity invisible under Ubuntu 14.04 and 14.10</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=83355">Bug 83355</a> - FTBFS: src/mesa/program/program_lexer.l:122:64: error: unknown type name 'YYSTYPE'</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=83380">Bug 83380</a> - Linking fails when not writing gl_Position.</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=83418">Bug 83418</a> - EU IV is incorrectly rendered after git1409011930.d571f2</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=84178">Bug 84178</a> - Big glamor regression in Xorg server 1.6.99.1 GIT: x11perf 1.5 Test: PutImage XY 500x500 Square</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=84355">Bug 84355</a> - texture2DProjLod and textureCubeLod are not supported when using GLES.</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=84529">Bug 84529</a> - [IVB bisected] glean fragProg1 CMP test failed</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=84538">Bug 84538</a> - lp_test_format.c:226:4: error: too few arguments to function ‘gallivm_create’</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=84557">Bug 84557</a> - [HSW] "Emit ELSE/ENDIF JIP with type D on Gen 7" causes Atomic Afterlife and GPU hangs</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=84651">Bug 84651</a> - Distorted graphics or black window when running Battle.net app on Intel hardware via wine</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=84662">Bug 84662</a> - Long pauses with Unreal demo Elemental on R9270X since : Always flush the HDP cache before submitting a CS to the GPU</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=84807">Bug 84807</a> - Build issue starting between bf4aecfb2acc8d0dc815105d2f36eccbc97c284b and a3e9582f09249ad27716ba82c7dfcee685b65d51</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=85189">Bug 85189</a> - llvm/invocation.cpp: In function 'void {anonymous}::optimize(llvm::Module*, unsigned int, const std::vector<llvm::Function*>&)': llvm/invocation.cpp:324:18: error: expected type-specifier</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=85267">Bug 85267</a> - vlc crashes with vdpau (Radeon 3850HD) [r600]</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=85377">Bug 85377</a> - lp_test_format failure with llvm-3.6</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=85425">Bug 85425</a> - [bisected] Compiler error in clip control operations in meta</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=85429">Bug 85429</a> - indirect.c:296: multiple definition of `__indirect_glNewList'</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=85454">Bug 85454</a> - Unigine Sanctuary with Wine crashes on Mesa Git</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=85647">Bug 85647</a> - Random radeonsi crashes with mesa 10.3.x</li>
<li><ahref="https://bugs.freedesktop.org/show_bug.cgi?id=85691">Bug 85691</a> - 'glsl: Drop constant 0.0 components from dot products.' broke piglit shaders/glsl-gnome-shell-dim-window and a few others with Gallium</li>
unsignedproperties[TGSI_PROPERTY_COUNT];/* index with TGSI_PROPERTY_ */
};
externvoid
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
